Everything worth knowing about spire

From sign-up to the REST API — the main modules, settings and keyboard shortcuts in one place. Scroll down, or jump anywhere from the table of contents on the left.

Getting started

Sign-up, your first workspace, your first project — in just a few minutes.

  1. Sign up

    Open spire.hu, click the Sign up button, and enter your name, email address and a password of at least 8 characters. You can also sign in with a Google account — no password needed, Google handles identification.

    After a password sign-up we send a confirmation email: its link verifies your email address. If it doesn't arrive, you can request it again from the app (at most 5 times an hour). With Google sign-in, an address verified by Google counts as verified. Accepting an invitation inside the app requires a verified email address.

  2. Create a workspace

    After signing up you'll land on the New workspace page. Here you set the workspace name, its URL (e.g. spire.hu/acme), time zone and currency.

    During the beta every spire feature is free — no card details required. The only limit is the monthly AI allowance (see AI credits).
  3. First project

    In the sidebar click Projects, then the New project button, and simply set the project's name, color and (optionally) icon. The project key (e.g. ACME) is generated automatically from the name and cannot be changed later.

  4. First task

    On the project page, use the Create task button or add one inline in the list / kanban view's chosen section. Tasks are identified by the project key and a running number: ACME-1, ACME-2, …

  5. Invite a teammate

    On the Settings → Members page you invite new members by email (with Admin / Member / Guest roles). The invitee joins via the link in the email; if they already have a spire account with a verified email address, they also see the invitation inside the app and can accept it there. Only workspace members can be added to projects and teams — one by one, or by adding a whole team (if any) at once.

If you leave during workspace creation

If you close the browser on the New workspace page without submitting the form, on your next login you land back on the same page. The system deliberately does not create a "default" workspace for you — the name is your call. If you change your mind, the Delete account link at the bottom of the page is there for you (right to erasure under GDPR).

Projects

When there are many tasks, projects keep the work organized and visible.

A project is a standalone unit of work inside your workspace — typically a product, a client, a domain or a larger initiative. Every task belongs to exactly one project. The ⭐ icon in the project header lets you favorite it — favorites appear in a dedicated sidebar block for quick access.

Statuses

StatusWhen to use it
PlannedDecided on, but not yet started
ActiveWork is in progress
PausedTemporarily halted, will continue later
DoneFinished, but still on hand
ArchivedClosed, reference only

Views

A project can be seen in three views — use the icons on the left of the Tasks tab toolbar, before the search input, to switch between them (desktop / tablet):

List

Tabular view. Grouped by status, one task per row. Keyboard-navigable with J/K, opened with Enter.

Kanban

Columns per status. Drag-and-drop to move tasks, status updates automatically.

Timeline

Gantt-like timeline view. A task is shown from its creation date to its due date, revealing overlaps and deadlines.

Sections

The List and Kanban views group tasks by status by default — using the grouping switch (next to the view switcher) you can switch to Custom sections. You can create any sections you like inside a project (e.g. "Quotes", "Orders", "Follow-up"), reorder them with drag-and-drop, and new tasks land in the lane where you create them.

Custom fields

Beyond the built-in fields (status, priority, assignee, due date, labels), inside a project you can create your own data fields with the + Custom field button — e.g. "Customer ID", "Budget", "Source". Eight types to choose from: text, number, date, checkbox, URL, email, single-select and multi-select list. The fields appear in the details panel of every task in the project and are fillable inline.

Tasks

Tasks are the atoms of spire — everything that needs to be done is a task.

A task is an independently executable item: a bug fix, a feature, a design decision, a to-do. Every task belongs to a project and optionally to a cycle, with exactly one assignee.

Types

TypeWhen to use it
TaskGeneral to-do — the default type
BugBroken behavior that needs to be fixed
RequestNew feature requested by a customer (external or internal)
IdeaA not-yet-finalized thought

Statuses

StatusMeaning
BacklogNot yet prioritized, waiting its turn
TodoPrioritized, ready to start
In progressSomeone is actively working on it
In reviewDone, but awaiting review
DoneCompleted and accepted
CancelledWon't be done

Priorities

PriorityWhen to use it
UrgentNeeds immediate attention
HighImportant, soon
MediumNormal priority
LowIf there's time, no rush
NoneUndecided / not relevant

Details panel

Clicking a task slides in a side panel from the right. Here you find…

  • title + description — title is plain text, description is rich text (formatting, links, images), both inline-editable.
  • properties — assignee, due date, type, status, priority, labels, all changeable inline from dropdowns.
  • project, section — the task's current project and section assignment.
  • custom fields — the values of the fields defined on the project (text / number / date / checkbox / URL / email / single- or multi-select list) are filled in here.
  • cycle + estimate — assignment to a running cycle and an estimated effort (hour = point).
  • sub-tasks — one level deep: a task can be broken into smaller to-dos, with aggregate progress. The ✨ Generate button in the Subtasks section header asks the AI to propose subtasks from the parent's description — you can edit each suggestion or uncheck it before finalising.
  • attachments — up to 50 MB per file: images (PNG, JPEG, GIF, WebP), PDF, text / CSV / Markdown / JSON, and Word, Excel and PowerPoint documents. SVG is rejected for security reasons. Attachments can only be downloaded while signed in, with access to the task.
  • comments + reactions — threaded comments with the rich-text editor, @mentions (sending a notification to the mentioned member; mentioning a team notifies every team member who can see the issue), and emoji reactions on every comment.
  • activity — chronological log: who changed what, when.

Undo

Most changes and deletions show a 5-second Undo toast (or ⌘+Z): deleting a task, sub-task, comment or attachment, status and assignee changes in the details panel and the command palette, kanban moves, archiving in the Inbox, deleting a label or team, and removing a member. Deleting a project asks for confirmation first (type the project name), then it can be undone the same way.

Cycles

Shorter, time-boxed iterations — spire's equivalent of a sprint.

A cycle is a time interval (typically 1–4 weeks) in which you work through a batch of tasks. A workspace can have any number of cycles, but only one can be active at a time.

Cycle lifecycle

StatusMeaning
PlannedCreated, but not yet started
ActiveCurrently running — highlighted on the cycles view
DoneClosed; statistics show performance

Assigning tasks

You can add a task to a cycle from the details panel or on the cycle page. The cycle page has three tabs: Overview (stats, burndown chart and AI summary), Tasks (every assigned task as a list), and Details (description, properties).

AI summary

Below the burndown on the Overview tab, an AI-generated narrative gives you an at-a-glance read on the cycle — pacing (where you are vs. where you should be), who is most active, what task types dominate (bug, feature, idea), deadline slippage, and a recommendation for the remaining time. Refresh with one click; the text is workspace-shared (everyone sees the same generated version, we don't burn tokens per viewer). At least 3 tasks must be in the cycle to generate, and it's refreshable once per day.

Retrospective (on closed cycles)

When a cycle closes — either its status changes to Done or its end date passes — a three-card retrospective form appears at the bottom of the Overview tab: "What went well?", "What didn't go well?", "What to improve?". 500 characters per card, auto-saves as you type. The team now has a permanent place to reflect on the cycle, and the lessons are at hand when planning the next one.

While the cycle is running, the AI summary is active and refreshable. When it closes, the last generated summary becomes read-only and the retrospective takes over — at any time one surface shows "current state", the other "lessons learned".

Time tracking

How much time went into an issue — with a timer or logged afterwards, plus billable hours.

Time tracking is for following work and billing — it is not a working-time record under the Hungarian Labour Code (Mt.) and does not replace one. spire takes no screenshots and does not measure keyboard or mouse activity: only what you log is recorded.

Timer

In the issue detail panel the Start button of the Time section starts the timer (also from the command palette: Start timer on this issue). You can have one running timer at a time — starting a new one stops the previous one. The running timer shows at the top of the sidebar (in the header on mobile) with the elapsed time; click it to open the issue, ■ stops it, and the bin discards it without saving (after a confirmation). Starting on the laptop and stopping on the phone works too: the server keeps the time.

A timer shorter than a minute is not saved. After 10 hours you get a notification; at 24 hours the timer stops by itself (with 24 hours).

Logging time manually

Add time lets you log time afterwards. The duration is free text: 1.5 (an hour and a half — a bare number means hours), 1:30, 1h 30m or 90m. The date can be up to a year back and one day ahead; a note and the Billable flag can go with it. You can edit and delete your own entries (undo within 5 seconds).

Weekly timesheet and report

Under Time in the sidebar, the Timesheet shows your week by issue and day: type into a cell to log, add rows with the issue search or copy last week's rows. The Report groups the period's hours by project, person, issue or day, and exports to CSV or Excel.

Hourly rate and invoicing (owner / admin)

Owners and admins can set an hourly rate on the project's Details tab, and whether new entries are billable by default; the report then shows amounts. Mark as invoiced first shows how many entries, how much time and what amount will be locked; the marked entries keep their rate and cannot be edited until released. If someone changed something in the meantime, nothing is locked and you get a fresh preview.

Who sees what

RoleLoggingVisibility
Owner / adminFor anyoneEveryone's time, with rates and amounts
MemberOnly their own, on an accessible issueTheir own entries, the total time of the issue and of their projects — never a rate or an amount
Guest—No access to time tracking

Inbox

The central place for your notifications — @mention, assignment, status change, comment.

The Inbox page (/inbox) shows every notification that concerns you — newest on top. Unread ones are marked in blue, and you can also see the count on the Inbox icon in the sidebar / bottom nav, updating in real time.

Notification types

TypeWhen you get it
AssignmentA task is assigned to you
Status changeThe status of a task assigned to you changes
New commentSomeone comments on your task
@mentionSomeone mentions you in a comment or description
InvitationYou're invited to a new workspace
Due dateYour task is due today or tomorrow

With the keyboard

ShortcutAction
JMove to next
KMove to previous
EnterOpen the related task
EArchive
Shift+EMark as read

Email notifications

Besides in-app notifications you can also receive email about the same events — each type can be toggled on/off on the Settings → Notifications page. In-app notifications are all on by default; among emails the comment, status change and @mention types are off by default to reduce noise, the rest (assignment, invitation) are on by default.

Push notifications

For the most important events we can also send a push notification to your device, even when spire is not open. To turn it on go to Settings → Notifications and press Turn on on the Push notifications on this device card at the top — the browser then asks for permission (it never asks by itself). The Send test button tries it right away. Which events send a push is set per type in the Push notifications section of the same page; by default assignment, @mention, invitation, due date, AI agent approval and a long-running timer are on, comment, status change and automation are off.

Each device is turned on separately (your phone and your laptop are listed under the card, and you can remove one from there). To turn it off press Turn off on the card, or revoke the permission in the browser / phone notification settings. After you sign out the device receives no more pushes.

iPhone / iPad: push only works in spire installed on the home screen (iOS 16.4 or newer) — in Safari: Share → Add to Home Screen, then open spire from the installed icon and turn it on there. In a browser tab the card shows the install steps. If you do not open the app for a week your sign-in expires and the push stops with it — turn it on again after signing in.

Insights and blockers

What got stuck, what is late, what blocks — and exactly who sees what.

The Insights page (/insights, below Home in the menu) checks tasks, projects and cycles every hour and flags a blocker when something is overdue, sits in the same status for too long, does not move, is blocked by a stuck task, is urgent but unassigned, was reopened several times, or when a cycle or project is projected to slip. Each blocker shows why (e.g. “5 workdays in In progress (threshold: 5)”) and a suggestion for what to do. You can open, snooze (until tomorrow, for a week, two weeks or a month) or dismiss it — undoable for 5 seconds. Workdays come from the workspace's work calendar.

Owners / admins switch signals off, tune the thresholds (one click back to the default) and set the estimate unit (points or hours) under Settings → Insights. A new high-severity blocker sends a notification (at most 5 a day), configurable under Settings → Notifications.

The Team tab shows the next two working weeks: a row per person, each day the estimated hours of the tasks due that day tinted against the day's capacity (under, near, over), before them the tasks in progress and this week's logged time against the weekly capacity; clicking a name opens that person's tasks. With point estimates a task counts as the configured number of hours. Owners / admins set the weekly capacity (40 hours by default, 0 = not plannable) underSettings → Members. The Time tab is the familiar time report plus an estimate vs. logged comparison with overruns highlighted. A project lead sees the tint of a colleague's days (measured against the colleague's capacity) but not the capacity or the logged time — wherever you may not see a number, a dash stands in its place.

The Projects tab shows how each project is doing: the computed health (On track, At risk, Off track — from the share of overdue tasks, the projected finish and high-severity blockers, with the reason), progress, the target date and the projected finish, the overdue share, open blockers and the estimated hours (logged time only for owners / admins). The project lead or an admin can write a status update on the project's Details tab: a chosen health (prefilled with the computed one, but yours to change) and a short text. The project's members are notified, and when the update says something other than the numbers, the Projects tab shows both. The author or an admin can delete an update (undo within 5 seconds); after 14 days without an update the project lead gets a reminder as a blocker.

Daily briefing

On workdays a short summary of your day is ready in the morning (07:30 by default, in your own time zone), waiting on a card at the top of the Home page: what you finished yesterday, how much time you logged per project, who wrote to you that you have not answered yet; what is due today, what is in progress with you, whether a cycle ends; and the blockers meant for you, each with its why. A project lead also sees their team (completed and in-progress tasks and blockers per person — in name order, never ranked) and the health of their projects; an owner / admin sees the company numbers too (completed and new tasks, total logged time, blockers by severity, the riskiest projects). On Mondays project leads and admins also get a weekly review. Every line is a live link to the task, the project or the filtered list.

Open takes you to /briefing, where you can page back 30 days; Refresh rebuilds today's briefing from the current data (at most 3 times a day). If the workspace and you have it switched on, the briefing starts with a short AI summary and up to 3 suggested steps; it comes from the workspace's AI credits (and counts toward your own monthly cap), and when they run out the briefing is made without it, with a sentence written from the numbers. Time, days, AI summary and weekly review: Settings → Notifications → Daily briefing; you can ask for email or push there too (by default it only appears in the app — on the card, Email it to me too is one click). The email carries only your own day; the team, logged time and the AI text are never in it. Nobody who has not signed in for 14 days gets a morning briefing.

On a blocker card, Why is it stuck? has the AI read the task's description, latest comments and history, and write the likely cause and the next step in two or three sentences. It runs only when you ask, from the workspace's AI credits. Colleagues' names go to the AI as aliases, email addresses and phone numbers never; the answer is only a suggestion.

Who sees what — employee notice (Hungarian Labour Code § 11/A)

Blockers show where work is stuck, not how people perform: there is no ranking, score or performance measurement. This is exactly what each role sees:

RoleSeesDoes not see
MemberBlockers on their own tasks (as assignee or creator), those their task blocks, blockers of the projects and cycles they can access, and their own personal signals (too many tasks in parallel, upcoming work over capacity, missing time entries), their own weekly capacity, their own workload (planned work per day against capacity) and their own activity (what they completed or started, how many comments they wrote, how much time they logged in a period)A colleague's task blockers, anyone else's personal signals, anyone else's weekly capacity, workload and activity
Project leadIn addition, the blockers of every task in the projects they lead (with the assignee's name, as on the task), and a colleague's “too many tasks in parallel” and “upcoming work over capacity” signals — only when the colleague is over the limit or the capacity counting the led projects alone, and the number (task count, planned hours) comes from those projects only; the workload of the colleagues working in the led projects (planned hours, tasks in progress — from those projects only) and a colleague's activity in the led projects (tasks completed, started, moved or created, the fact of comments — not their text)Since when the colleague has been over the limit, the colleague's weekly capacity (it can be inferred from the colours of the workload cells), work and activity in other projects, logged time, missing time entries
Owner / adminEvery blocker, every member's “too many tasks in parallel” and “upcoming work over capacity” signals, every member's weekly capacity (which they set), every member's workload and activity including logged time (activity search never shows the text of comments), including facts derived from logged timeAnother member's “missing time entries” signal (only the person sees it)
Guest—The Insights page

Who can act on a blocker: on a task the assignee, the project lead or an admin; on a project or cycle the project lead or an admin; on a personal signal only the person or an admin. Every snooze, dismissal and reopening goes into the audit log (who, when, which signal), as does a change of project lead. When a project lead or an admin searches a colleague's activity, that goes into the audit log too (who, about whom, which period — without the results). An external AI assistant (MCP) or an API token receives another person's personal signals, workload or activity only when the token was explicitly granted the team insights permission (none is by default), and even then only what the role sees according to the table above.

The briefing follows the same rules, and you only ever see your own — nobody sees anyone else's, admins included. A member's briefing holds only their own work. In a project lead's team section a colleague shows the number of completed and in-progress tasks (from the led projects only) and the blockers of the colleague's tasks; it does not show the colleague's workload, logged time or personal signals (too many parallel tasks, capacity, missing time). An owner / admin's team section adds yesterday's logged time and the workload, and how many people logged no time — a number only, no names. If you lose access to something in the meantime, the briefing shows “Not available” in its place. Briefings are deleted after 90 days.

Quick capture

Say it freely or type it — spire AI turns it into a structured task.

The Quick capture page (/quick-capture) opens on desktop from the navigation results of the Command palette (⌘K, or the sidebar search field); on mobile the middle + button in the bottom nav leads there. Type or speak anything (up to 8,000 characters) — the system parses it and turns it into a task proposal that you can still tweak before confirming. Quick capture is not available to guests.

What it understands

  • Task name: the to-do extracted from the text (e.g. "bug fix for login")
  • Project: if you mention one by name or key
  • Assignee: name or first name ("Tibor", "Monika") → workspace member
  • Status: e.g. "started", "in progress", "done" → status value
  • Priority: "urgent", "important", "someday"
  • Due date: "tomorrow", "next Monday", "in 2 days", "by Friday"

Voice capture

Click the microphone icon inside the input to record speech — through the browser's built-in Web Speech API. spire's server never receives audio, but the browser runs speech recognition with its own service: in Chrome the audio goes to Google's speech recognition service. The recognized text is then processed the same way as typing.

Example: "tomorrow Monika contract prep for ABC Ltd, urgent" — the system parses this as: assignee = Monika, due date = tomorrow, priority = Urgent.

Batch capture

You can capture multiple tasks at once — in the generated list you can edit, delete or skip any of them before saving. Meeting notes turn into a task list in seconds.

AI allowance

Quick capture uses the workspace's shared AI credits (see AI credits). On top of that the limit is 500 AI calls per user per month and 15 requests per minute. On the Quick capture page a warning bar appears from 80% of the allowance; once it is used up, AI generation pauses and manual capture keeps working the same way. Failed AI calls don't count toward the call count.

For AI processing, Quick capture sends the typed text, the workspace members' names and email addresses, and the list of projects to an external AI provider (Claude on Amazon Bedrock, from a region inside the EU — Frankfurt by default — with processing only in EU member-state regions and no model training). Details in section 6 of the Privacy Policy.

Collaboration

Workspace-level roles, teams and invitations.

Roles

RoleWhat they can do
OwnerEverything an admin can do + workspace deletion and ownership transfer (both require re-authentication). Each workspace has exactly one owner.
AdminWorkspace settings, member management, labels, teams, webhooks; sees and can revoke every member's API tokens and connected apps.
MemberSees public projects and the private projects they are a member of; creates and edits projects and tasks there, comments. Can create their own API tokens and connect AI assistants, always with their own permissions. No access to workspace settings (their own settings, yes).
GuestOnly sees projects they are explicitly a member of (not even public ones). Can comment, react, and change the status of tasks assigned to them. Cannot create projects or tasks, use Quick capture, delete, run bulk actions, create API tokens or connect AI assistants.

Invitations

On the Settings → Members page you enter an email address and a role. The invitee gets an email with a link valid for 7 days that leads to a preview page: they see which workspace they were invited to and on which email address. If they don't have a spire account yet, they continue from there to sign up (email pre-filled); if they do, they log in. Then a single Join button adds them to the workspace with the given role. Guests can be invited too — they only see the projects you add them to afterwards.

If the invited address already belongs to a spire account with a verified email, the pending invitations also show up in the app (a bar at the top and a list on the /workspaces page) and can be accepted there without the link. Within an hour at most 30 single and 5 bulk (up to 50 addresses each) invitations can be sent.

Teams

Within a workspace you can organize members into teams — for example "Marketing", "Sales", "Customer support". A member can belong to multiple teams. You can also add a whole team to a project in one go (the team's members are individually added to the project members).

Settings

Personalizing your account and workspace — one quick overview per tab.

Personal settings

Profile

Name, job title, avatar, email address, interface language (Hungarian / English / Russian; this help is available in Hungarian and English, the legal documents in Hungarian), time zone — these only concern you. With a Google account, the Sync Google profile picture switch keeps your avatar in step with your Google picture; uploading your own picture turns it off. This is also where you find the GDPR data export and account deletion. Deletion requires re-authentication (password, a fresh Google sign-in for Google accounts, plus a code when 2FA is on) and typing the confirmation word. While you own a workspace that has other members, the account cannot be deleted — transfer ownership or delete the workspace first; workspaces used only by you are deleted together with the account.

Notifications

Per-type on/off toggles for email and in-app notifications. You decide which events send notifications.

Security

Change password (signs out your other devices), set a first password for a Google account, connect / disconnect a Google account, two-factor authentication (TOTP) with 8 single-use recovery codes — when enabled, a required second step at sign-in. Enabling 2FA, connecting / disconnecting Google and setting the first password require re-authentication. List of your last 50 sign-ins (view only).

Workspace settings (owner + admin)

Workspace

Name, URL (slug), time zone, currency, logo upload. Ownership transfer and workspace deletion are owner-only and require re-authentication; deletion also requires typing the exact workspace name.

Members

Invite by email, change role (owner / admin / member / guest), remove members; clicking a member opens the member panel with their module permissions too (see Modules and permissions).

Modules

Switch business modules on and off for the workspace; switching off keeps the data (see Modules and permissions).

Teams

Grouping members inside the workspace (e.g. Marketing, Sales, Customer support). A team can be added to a project in one step.

Labels

Workspace-level label library (color, name) — labels are assignable to tasks.

API tokens

Generate tokens for programmatic access with a name and optional expiry. A token created in the UI is unrestricted (it gets its creator's workspace permissions); narrower scopes can only be set through the API (see REST API).

Import / Export

Full workspace JSON export (backup / migration) and JSON import. Task import from CSV, Excel (.xlsx) or a Trello export (JSON), up to 10 MB and 5,000 tasks: the wizard recognises Jira, Asana, Trello and Hungarian headers (AI can help with the mapping on request — only the headers and a few sample rows, with e-mails masked), and the preview shows what will be created and which rows have errors (downloadable error CSV). Comments arrive with their original author and time, attachments as links. The import runs no automations and sends no notifications; it can be undone for 7 days from the “Previous imports” list — anything changed since then is kept.

Audit log

Who did what and when inside the workspace. Searchable, filterable. Visible to owner + admin.

Billing

Pricing overview; for the owner the active plan, monthly / annual cycle switching and the issued invoices (PDF). Free during the beta, no invoices are issued; after the beta invoicing is (for now) manual.

Modules and permissions

Business modules (CRM, invoicing, HR, calendar …) next to task management — switched on per workspace, with access controlled per member.

Modules are arriving gradually. The framework is in place, but right now no module can be switched on yet, so the Settings → Modules page is still empty. As each module is finished it shows up there — first in a few workspaces by invitation, then for everyone.

Switching modules on (owner + admin)

The Settings → Modules page has one switch per module, grouped (customers and sales, finance, operations, people, collaboration, AI). Each row says what the module is for, what it requires (e.g. finance needs invoicing), and whether it handles personal or sensitive data. If a required module is still off, the Turn this on too button turns them on in order.

Switching a module off keeps its data — it only disappears from the interface, and when you switch it back on everything is where you left it. If another module depends on it, we ask for confirmation first (the dependent module becomes inactive too), and switching off can be undone for 5 seconds. Active modules appear in the Modules section of the left sidebar and in the command palette (Go to: …).

Module permissions per member

On the Settings → Members list, clicking a member opens the member panel; on its Module permissions tab you set what they can reach in each active module:

LevelMeaning
No accessThe module does not exist for them: hidden from the menu, its address shows a "not available" page.
ViewRead (in some modules, e.g. HR, only their own records).
EditCreate, edit, delete their own records.
ManageThe module's settings, deleting others' records, import, export.

Where you set nothing, the role's default applies (the Default option shows what it is), and some modules add separate capabilities (e.g. issuing invoices). The rules: the owner always has full access, guests cannot reach modules; nobody can raise their own permissions; only the owner can change an admin's module permissions; and nobody can grant more than they have themselves. A change takes effect immediately and is recorded in the audit log.

AI agents and AI assistants

An AI agent is a member too, so the same Module permissions section in its edit dialog narrows what it can reach. During a run the agent's permissions are intersected with those of the requesting human — an agent never sees or does more than the person it works for. The tools of a module that is switched off are available neither to agents nor to your own AI assistant (MCP); the list_modules tool tells the assistant which modules you can use.

Company profile and work calendar

Your company details that go on documents, and the work calendar that tells spire which days are workdays.

Company details (owner + admin)

On Settings → Company details you enter the legal name, the kind of business, the tax number, the EU VAT number, the registration number, the address, the contact details, up to five bank accounts (one of them the default), the VAT status and the default payment terms. You save section by section; spire checks the tax number and the bank account number (domestic number or IBAN) by their check digits. While the address or — for a company — the tax number is missing, a banner tells you what documents still need. On an empty page the owner can fill it in from the billing details with one click (nothing is saved until you press save). The upcoming modules (quotes, invoices, work orders) build on these details.

Bank account notice. When someone adds, removes or changes the default bank account, the owner gets an email about it (even when they made the change) — so an unauthorized change does not go unnoticed. Members can only view the details, with bank accounts masked; guests cannot see them.

Work calendar

On Settings → Work calendar you set which weekdays are workdays, and the exceptions live here: Hungarian public holidays, transferred working and rest days, and your own company days off or make-up Saturdays. A new workspace starts with this year's and next year's official calendar; you can load or refresh a year with Load official calendar — you see the changes first, and your own exceptions are never touched. If the transferred days of a year are not published yet, only the statutory holidays are loaded. Deleting an exception can be undone for 5 seconds. Every member and guest of the workspace sees the name of an exception, so keep personal details out of it.

The due-date picker of tasks marks the calendar: a holiday is dimmed and shows its name on hover, a transferred working day gets a dot — the day can still be picked. Your AI assistant counts in workdays with the get_work_calendar tool (e.g. "what is 5 workdays after Friday?"); the get_company_profile tool returns the company profile only when you explicitly granted it to the token.

Partner register

Customers, suppliers and their contacts in one place — every module that works with customers reads from the same register.

The partner register is not a module of its own. The Partners and Contacts entries appear in the sidebar when the workspace has a module that uses the register (e.g. CRM) switched on and you have at least view access to it. Without such a module, or as a guest, the pages are not available.

Partners and contacts

On the Partners page you keep companies, sole traders, private customers and public bodies, with roles (customer, supplier, prospect, business partner), a tax number, addresses, bank accounts, an owner and tags. Spire checks tax numbers and bank accounts (a wrong check digit gives a warning, saving still works). Search finds names without accents and with typos too (“kovacs”, “kovcs” → Kovács). Contacts are people who work at one or more organisations, with their own details and data-handling flags (legal basis, marketing consent, do-not-contact).

When you save a new partner, spire warns if a record with a similar name, tax number, e-mail or website already exists — this is a hint, not a block (one company can have two shops). A live partner with the same tax number cannot be created twice, though.

Timeline and activities

On a partner or contact panel the Timeline tab shows notes, calls, meetings, e-mails, the record's changes and files in one stream. A call shows on the contact and on its organisation. On a phone, after you tap call and return to the app, spire offers to log the call.

Archiving, deleting, personal data

A partner can be archived (reversible, it disappears from pickers). If a module references it, it cannot be deleted, only archived. Data of private customers and sole traders can be exported and, on request, anonymized (confirmed by typing the name): personal fields are erased, the accounting reference stays. Deleting, exporting and anonymizing are only available in the app, with your own session.

Products and services

One shared catalog of products, services and fees — sales, invoicing and inventory all pick items from it.

The catalog is not a module of its own. The Products and services entry appears when a module that uses it (e.g. CRM, sales, invoicing or inventory) is switched on and you have at least view access to it. Adding and editing items takes the sales or the inventory module; with the other modules the catalog is read-only.

Items

Every item has a type (service, product or fee — e.g. a call-out charge), a unit (piece, hour, day, km, kg, litre … or your own, e.g. “set”), a net list price in whole forints and a VAT code (27 %, 18 %, 5 %, 0 %, AAM, TAM, EU, EUK and the other exemption codes). You can also type the price gross: spire works out and stores the net amount and shows the gross next to it. A new item's VAT code starts from your company details (AAM for a VAT-exempt small business, 27 % otherwise). An SKU and a barcode can be added; both are unique within the workspace, and spire warns about a barcode with a wrong check digit. Search finds items by name, SKU and barcode, without accents and with typos too, an exact SKU first. Items can be filed in a three-level category tree (Settings → Partners and catalog).

Purchase price

An item's purchase price is the other half of your margin, so only the owner, admins and members with access to the inventory or finance module see and edit it. An AI assistant or an API token does not get it even if its owner could see it — only with a separate catalog-costs scope granted for exactly that.

Deactivating and deleting

An item you no longer sell can be deactivated (reversible): it cannot be picked for a new quote or invoice, but old documents still point at it. A deleted item can be restored for 30 days, then it is removed for good; while a document references it, it cannot be deleted, only deactivated.

Documents

Numbered, branded PDF documents — quotes, contracts, delivery notes and the rest — that your customer opens and accepts from a link.

Documents are not a module of their own. A document is always issued by a module (sales issues the quote, for example), and whoever can access that module can access its documents. The Settings → Documents page is for the owner and the admins, and appears only when at least one module that issues documents is switched on. Invoices are not numbered here — your invoicing provider (Billingo) issues them.

Numbering

Every document type has a gapless series that restarts every year, e.g. AJ-2026-0001, AJ-2026-0002. Only an issued document gets a number, a draft does not; a new version of the same document keeps the number with a -v2 suffix. An issued document cannot be deleted, only voided — the number and the reason stay in the Number log, which you can export to CSV (for your accountant, say). You set the prefix and the pattern on the Series tab; once a number was issued in a year, a change applies from the next year. If you move to spire mid-year, you can raise the starting number.

Branding and templates

On the Branding tab you pick a layout (classic or modern) and an accent colour, and set whether your logo shows, what goes in the footer, the default language and how many days a link lives. A live PDF preview on the right shows the change before you save. On the Document templates tab you write your own texts per type (introduction, terms), reorder the blocks by dragging, and insert fields (the recipient's name or the number, for example). The issuer's details come from the Company profile: with an incomplete profile no document can be issued. spire prints the mandatory VAT wording (reverse charge, VAT exemption …) itself from the lines' VAT codes. The PDF of an issued document never changes — a change to the company profile or a template only affects documents issued afterwards.

Sharing by link

You can create a link to a document for viewing or for acceptance; you can copy it again at any time, extend it or revoke it (a revocation can be undone for 5 seconds). The recipient needs no account: the link shows the document in a readable form, the PDF can be downloaded, and on an acceptance link they accept or decline with their name and a checkbox. The link always shows the latest version. You are notified when it is first opened and when a decision is made — an opening counts only when someone actually loaded the page (not a mail client's automatic link check), and not your own signed-in opening. The notification names the document type and number, not the customer; you see the details on the document's panel.

What an acceptance means legally

An acceptance given on the link is a simple electronic signature: under the eIDAS Regulation its legal effect cannot be denied solely because it is electronic, and under the Hungarian Civil Code the written form is met when the statement can be retrieved unchanged and the person and the time of the statement can be identified. That is why spire records the document's unchanged PDF and its fingerprint (SHA-256), the name of the person accepting, their IP address, their browser and the time.

This is not a qualified electronic signature. The law may require a qualified electronic signature or a public deed for some contracts; an acceptance in spire does not replace those.

Approvals

One place for what waits for your decision and what happened to your own requests.

On the Inbox page, the Awaiting approval tab collects the requests waiting for your decision; the number on the tab shows how many you can decide right now. In the Waiting for me view you approve or reject (a rejection can carry a reason the requester will see), and in My requests you follow what happened to yours. When a decision can be taken back, the Undo button gives you 5 seconds to change your mind; when it cannot (for example because it has an effect outside spire), spire asks you to confirm first.

Today the tab shows AI agent approvals (when an agent asks before deleting something). The shared approval framework is ready: upcoming modules (discount, leave or invoice approval, for example) will bring their requests here. You cannot approve your own request unless nobody else can decide it (for example the owner of a one-person company). You are notified in the app and by push (by e-mail only if you turn it on in the "Approvals" row of Settings → Notifications); e-mail and push leave out the details of the request — you see them in the inbox.

Templates

Starting from a ready structure is faster: a new project is set up with sections and starter tasks in a minute, a common task with its subtasks in one click.

There are two kinds. A project template is a whole project skeleton: sections, custom fields, labels, starter tasks (with subtasks and due dates relative to the start date), the project's task templates, and its automation rules and recurring issues. A task template is a prefilled new issue: title, description, status, priority, type, estimate, labels, custom fields, up to 50 subtasks and a due date relative to the creation day. A task template can serve the whole workspace or a single project.

Templates hold no people: assignees, watchers, teams, comments, attachments and history never go into one, and @mentions and uploaded images are removed from descriptions.

New project from a template

  1. Projects → Create project

    The first step lets you choose: an empty project, a built-in template, or your workspace's own project templates. The selected template's structure is previewed in the dialog.
  2. Name, key, start date

    Due dates follow the start date (today by default); color and visibility come from the template and can be changed.
  3. Create from template

    The whole project is created in one step, without per-task notifications. If something could not come over, the dialog lists it before you open the project.

Ten built-in starter templates are written for Hungarian small businesses: Agency — campaign, Construction — building project, Car dealership — vehicle preparation, Software — sprint, Event planning, New hire onboarding, Accounting firm — month-end close, Online shop — product upload and campaign, Service & maintenance and Client onboarding. The project is created in your language (Hungarian, English or Russian).

Automation rules from a template arrive disabled and recurring issues paused — review them before you switch them on. A rule badged “Needs review” lost a part (e.g. a condition on a person); it can be enabled only after the rule is saved. Rules are created only when an owner or admin creates the project — for a member they are left out, and an admin can take them over.

Issue from a template

Use the template picker at the top of the new-issue panel, the template button next to New issue on a project page, or New issue from template in the ⌘+K menu. The panel opens prefilled — change anything, drop subtasks or labels — and saving creates the issue together with its subtasks. If the project has a default task template, the new-issue panel opens with it; if you do not touch it, closing the panel creates nothing (⌘+Enter does). A workspace-wide template fills a custom field only if the project has a field with the same name and type — otherwise the panel tells you.

Save as template and managing templates

On any project's Details tab, Save as template takes a snapshot of the project's structure and, if you ask, of its open or all tasks; later changes to the project do not affect the template. An existing project template can be replaced the same way. A template saved from a private project becomes visible to every non-guest member of the workspace — the dialog warns you. On Settings → Templates you edit task templates, rename and delete project templates, and make an editable copy of a built-in one. A delete can be undone for 5 seconds; a project's default task template is set on its Details tab.

RoleWhat they can do
Owner, adminCreate, edit and delete templates, save as template, set the default task template
MemberUse templates: new project and new issue from a template (only the names of a project template's rules are visible)
GuestSees no templates

Creating from a template needs a connection — it is never put in the offline queue. A workspace can hold up to 200 templates (built-ins do not count).

Automations

Repetitive work without code: when something happens and the conditions hold, Spire does it for you.

A rule has three parts: WHEN (what starts it), CONDITIONS (when it should run) and THEN (what it does). Example: "when the status becomes Done → close every open subtask too and notify the project lead". Rules are managed by the owner and the admins under Settings → Automations; members and guests do not see the menu item. A rule applies to every project or only to selected ones. With an empty list you can start from 12 ready-made templates.

The actions are performed by a hidden "Automation" actor: the history and the comments show the rule's name with an "Automation" badge. A rule runs with the permissions of its owner (who created it, edited it last or enabled it) — if the owner is no longer an admin or cannot reach the project, the rule is switched off and the owners and admins are notified. Rules run in the background and never slow down your own actions.

The rule's name is visible to everyone who can see the issue (guests included), so do not give it a confidential name.

Editor: Flow and List view

You can edit a rule in two views; the switch sits below the rule details (name, description, projects). The Flow view (the default) shows the rule top to bottom as connected cards: WHEN → conditions → steps, each with a one-line summary. Use the “+” button between cards to insert a step or a branch, and the card’s ⋯ menu to move it up or down, duplicate or delete it; you can also reorder by dragging (mouse, touch and keyboard). Clicking a card opens the step in a side panel. A branch’s paths are shown side by side as columns on desktop and one at a time, switched with a chip row, on a phone; an invalid card gets a red border. The List view shows the same rule as a single form. Both views edit the same rule, and the browser remembers the view you picked.

Triggers (WHEN)

GroupFires when
Issueis created, is updated, its status / priority / type / assignee / labels / due date / cycle / a custom field changes, is moved, deleted, restored, all of its subtasks are done
Commenta comment is added (can be filtered by the author's role, e.g. guests only)
Timethe due date approaches (N hours / days), the issue becomes overdue, the issue has not changed for N days, on a schedule (daily, weekly, monthly or cron — in the workspace time zone)
Cyclea cycle starts / completes
Manualonly from the "Run now" button

Time-based triggers are checked every 5 minutes, and a rule normally fires only once for the same issue with the same due date (exceptions: the stale reminder can repeat, and an overdue / due-soon rule can fire again for an issue that stays unchanged for a very long time). Deleting a cycle or a section and importing data do not start rules.

Conditions

Conditions can be grouped with AND (all must hold) or OR (at least one must hold), up to 3 levels deep and 50 conditions in total. You can check the issue's project, section, cycle, parent, status, priority, type, assignee, creator, labels, due date, estimate, title and description, the state of its subtasks, the project's custom fields, the value before and after the change, and who caused the change (a member or a role). Conditions always look at the issue's current state. No conditions = every event matches.

Actions and branching (THEN)

Steps run in order (up to 20): status, priority, type, assignee (a given member, the person who made the change, the creator, the project lead or round robin from a team), add / remove / set labels, due date (relative or fixed), estimate, cycle (active, next, a given one), move, custom field, comment, create a subtask or a new issue, update the parent or the subtasks, notification (in-app, plus email according to the recipient's notification settings), webhook and stop (skip the remaining steps). With a branch you can add IF / ELSE IF / ELSE paths (up to 5 paths, 2 levels deep) — the first matching path runs. If one step fails, the others still run.

Texts can use variables, e.g. {{issue.title}}, {{issue.link}}, {{issue.assignee}}, {{actor.name}}, {{change.from}}, {{change.to}}, {{today}} — each also has a Hungarian name (e.g. {{feladat.cím}}). The editor lists all of them.

AI actions

AI comment (following your instruction, e.g. "summarise the issue and suggest a next step"), AI labelling (it can only pick from the labels you list), AI priority and AI subtasks (up to 7, never repeating existing subtasks). AI calls come from the workspace's AI credits (the rule owner's own cap does not stop them); when the credits run out the step is skipped but the rule stays on.

Webhook and signature verification

The webhook step sends a POST request to the given address (in production only a public https URL, at most 10 seconds, no redirects). The body: { event: "automation.rule", deliveryId, timestamp, data: { rule, trigger, eventType, actorId, issue } }. Headers: X-Spire-Signature (sha256= + the HMAC-SHA256 of the raw body with the workspace's automation secret), X-Spire-Event, X-Spire-Delivery (the same on retries — deduplicate on it) and X-Spire-Automation-Rule. The secret can be revealed and rotated on the webhook step of the editor; rotation takes effect immediately, with no grace period. Do not put passwords or tokens into the webhook URL.

The step does not wait for the receiver: the request goes into a delivery queue and is sent from there. If the receiver is unreachable, times out, or answers 408 / 429 / 5xx, Spire retries after 5 seconds, 30 seconds, 2 minutes, 10 minutes and 1 hour (at most 6 attempts); any other response (e.g. 4xx, a redirect) is a permanent failure. Deliveries run in parallel and with retries, so the order is not guaranteed and, rarely, the same delivery can arrive twice — always deduplicate on the X-Spire-Delivery header. A failed delivery does not switch the rule off; at most 1,000 deliveries can wait per workspace, above that the webhook step is skipped.

If the receiver sends a Retry-After header (in seconds or as an HTTP date), Spire waits at least that long before the next attempt — at most 1 hour, the number of attempts does not change. If an endpoint (the same host and path) fails 10 deliveries in a row, Spire pauses it: nothing is sent to it for 5 minutes, and waiting deliveries show as Paused without using up an attempt. After the pause a single probe delivery goes out; if it succeeds everything resumes, if not the pause doubles (at most 1 hour). The rule owner is notified when an endpoint is paused (at most once a day per endpoint). A delivery can wait at most 72 hours from its creation; after that it is closed as expired without being sent.

javascript
import crypto from 'node:crypto' // call it with the raw (not yet parsed) request body function verifySpireSignature(rawBody, signatureHeader, secret) { const expected = 'sha256=' + crypto.createHmac('sha256', secret).update(rawBody).digest('hex') const a = Buffer.from(signatureHeader ?? '') const b = Buffer.from(expected) return a.length === b.length && crypto.timingSafeEqual(a, b) }

Delivery log and redelivery

On the page of a rule with a webhook, the Deliveries tab (and the expanded run) lists the deliveries: status (pending, sending, delivered, failed), the receiver's response, the number of attempts and the time of the next one; only the host and path of the address are shown, never the body that was sent. A finished delivery can be sent again with Redeliver — with the same X-Spire-Delivery id and the same body, signed with the current secret (at most 30 per hour per workspace). Finished deliveries are kept for 30 days.

History (versions)

Every save of a rule's content (name, description, scope, trigger, conditions, steps) is a new version. The History tab shows who changed what and when; a version can be viewed, compared with the current one and restored (a restore is a new version too, so nothing is lost). The last 100 versions are kept per rule. If the rule was changed elsewhere in the meantime (another tab or another admin), the editor does not overwrite it: it shows a notice, and Reload loads the latest version.

Dry run and log

Dry run tries even an unsaved rule on the latest matching events or issues (up to 50) and shows step by step what would happen — it changes nothing and calls no AI. A manual or scheduled rule can be started right away with Run now (after a confirmation, on up to 500 issues). The Run log tab lists every run (success, partial, failed, skipped) with expandable step details. A rule that fails 5 times in a row switches itself off and its owner is notified.

The Notify on error setting decides per rule whether Spire tells the owner about a failed run: never, on the first error (default; not repeated until the next successful run) or on every error (at most 10 per hour). The notice links to the rule's run log. If rules triggering each other would do too much work (at most 500 automation-caused actions from one originating event), the chain stops and the owner of the busiest rule is warned; the rule causing the loop is switched off only if it runs away for the third time within 24 hours.

Limits

WhatLimit
Conditions3 levels deep, 50 in total
Steps / branching20 steps; 2 levels, 1–5 paths + else per branch
Execution1,000 actions / minute per workspace (above that, runs are delayed)
Chainingrules triggering each other at most 5 levels deep; a rule runs once per issue in one chain; at most 500 automation-caused actions per chain
Webhook deliveryat most 6 attempts (5 s, 30 s, 2 min, 10 min, 1 h); at most 1,000 pending deliveries per workspace; log kept for 30 days
Manual and scheduled runsup to 500 issues per run
Dry runup to 50 items
Log30 days, at most 1,000 runs per rule
A comment written by an automation cannot be edited; the owner and admins can delete it (with a 5-second undo). If a rule's comment mentions someone, the mentioned member gets access to the project, just like with a manual comment.

AI credits

spire's AI features draw on the workspace's shared AI credits: a monthly allowance per paid member, topped up with credit packs when needed.

What uses credits: every AI feature spire runs — Quick capture, sub-task suggestions, the cycle summary, import mapping, AI steps in automations, AI agents, the briefing summary and the “Why is it stuck?” analysis. Your own connected AI assistant (MCP, see AI assistant) works with its own model and uses no credits.

Metering: 1 AI credit = 1,000 AI tokens. An AI token measures what a call really costs in a fixed unit, so on the fast model an average call is roughly as many AI tokens as the text tokens it processes, while the smart model or a long answer uses proportionally more. spire never charges more than a call is actually worth.

Monthly allowance: 2,000 AI credits per paid member per month, in a pool shared by the workspace; it restarts at the beginning of the month and unused credits do not carry over. During the beta the allowance is free.

Credit packs: when the allowance is not enough, the owner can order a credit pack under Settings → AI usage; a pack is used after the monthly allowance and is valid for 12 months from payment. The order is invoiced and the credits are added when the invoice is paid; an unpaid order can be withdrawn. During the beta purchases are paused.

Alerts: at 80% the owner and those who manage AI get a notification (with a suggested pack); when the credits run out the AI features pause with a clear message — everything else keeps working and there is no overage invoice. A small indicator next to the AI buttons shows when credits run low.

Fair use: one person's own AI calls can use up to 6,000 credits of the shared pool per month by default (three times the per-member allowance); the owner or an admin can set this between 500 and 100,000 credits or switch it off. Agent and automation runs are limited only by the shared pool.

Who sees what: under Settings → AI usage the owner and admins see the balance, the forecast, the usage breakdown (by module, feature, person and agent) and the credit ledger; other members see their own monthly usage and the state of the workspace pool, without amounts.

AI agents

Your own AI teammates in the workspace: assign them an issue or @mention them, and they work in the issue's project — never with more rights than the person who asked.

AI agents are part of the AI module and cost nothing extra: their runs use the workspace's AI credits. The owner or an admin creates agents under Settings → AI agents: name, role, instructions, which tools it may use (e.g. read issues, comment, update), which projects it works in, and whom it may hand work over to. An agent is not a seat, cannot sign in, and is shown with an “AI” badge.

  1. Start

    Make the agent the assignee of an issue, or @mention it in the description or a comment. Only members (not guests) can start an agent, and only in a project they can write. Automations start it with the Run AI agent step (with the rule owner’s permissions).

  2. Work and activity

    The agent reads and writes in the issue’s project and reports back in a comment. The Agent activity tab of the issue panel shows what it does, step by step, and lets you stop it. An agent can hand the issue over to a teammate (at most 3 levels deep).

  3. Approval and undo

    For a deletion the agent asks for approval: the requester and the admins are notified and can approve or reject it on the issue panel or in the inbox (within 24 hours). Every agent write can be undone step by step if the issue has not changed since.

What the agent reads (the issue text and its comments) is sent to the AI provider configured by spire for processing. Run step summaries are kept for 30 days; prompt and reply texts are not stored.
Limits: at most 8 turns and 12 tool calls per run, 2 concurrent runs per workspace, as long as the workspace has AI credits. With the AI module off agents do not run.

Command palette

Fast navigation, search and AI in a single input — ⌘K.

The Command palette is spire's "one key for everything" surface. It opens with ⌘+K; on desktop by clicking the sidebar search, on mobile via the search icon in the bottom nav (next to the +). Type to search the current workspace's tasks, projects, cycles and members (only the ones you have access to) — status dots, priority icons and avatars are shown next to the results. The Quick capture page is among the navigation results.

Navigating inside the palette

ShortcutAction
↑ / ↓Move between results
EnterOpen the selected result
EscClose palette (or one level back)

Actions from the palette

On the right edge of task rows in the search a caret (›) icon switches to drill-in mode: status, priority, assignee and label can be changed directly from the palette, without opening the detailed view. After the mutation runs, a short toast confirms success, the palette closes, and list / kanban / timeline update everywhere instantly. Esc in the picker steps one level back (to the actions), two Esc close it entirely.

Keyboard shortcuts

spire is fully keyboard-navigable. These are the most common ones.

Global

ShortcutAction
⌘+KCommand palette / search
⌘+ZUndo last action (5-second window)
G+HHome
G+NInbox
G+PProjects
G+IMy tasks
G+CCycles
G+SSettings
⌘+/Help
?Help (alt.)

In lists

ShortcutAction
JNext item
KPrevious item
CCreate task (on project / tasks / cycle pages)
EnterOpen selected item
EscClose panel / clear focus

When editing a task

ShortcutAction
⌘+EnterSave (comment, new task, edit)
EscDiscard edit

In the inbox

ShortcutAction
EArchive selected notification
Shift+EMark selected notification as read
Shortcuts are only active when focus is not inside a text input. The full list can be pulled up at any time with ⌘+/.

REST API

Programmatic access — integrations, automation, your own client.

spire's REST API covers the workspace data (projects, tasks, cycles, labels, …). API tokens are generated per workspace on the Settings → API tokens page — every member except guests, each with their own permissions (owners and admins see and can revoke every token). Account-level operations (profile, password, 2FA, GDPR, workspace management) are not reachable with an API token — see below.

Basics

Base URLhttps://spire.hu/api (production) / http://localhost:3001/api (dev)
FormatJSON request body + JSON response
AuthenticationAuthorization: Bearer <token>
Rate limit120 requests per minute per API token, across all /api/* endpoints together; on top of that, per IP address and endpoint on sensitive endpoints (see the table). On overrun you get a 429 response with a Retry-After header.

Endpoint groups

Tenant-scoped (workspace-bound) endpoints are under the /api/v/ prefix; auth- and user-level endpoints are under /api/ and /auth/.

Auth- and user-level (no workspace context)

PrefixContents
/authLogin (+ 2FA step), sign-up, email verification, refresh, Google OAuth, password reset
/api/workspacesWorkspace list, create, switch, transfer, delete
/api/profileOwn profile, password, Google account, 2FA, onboarding flag
/api/gdprData export, account deletion, sign-in history
/api/lookupCross-workspace resource resolution (?type=project&key=ACME)
/api/invitationsInvitation preview + acceptance, own pending invitations

Workspace-level (/api/v/)

PrefixContents
/api/v/projectsProject CRUD, member and team assignment
/api/v/issuesTask CRUD, comments, reactions, bulk operations
/api/v/cyclesCycle CRUD, task assignment
/api/v/labelsLabel CRUD
/api/v/teamsTeam CRUD, member assignment
/api/v/membersWorkspace members, invitation, role
/api/v/notificationsNotifications, read state, preferences
/api/v/searchSearch across tasks, projects, cycles and members
/api/v/aiQuick Capture (triage), sub-task suggestions, import mapping suggestions, AI allowance status
/api/v/custom-fieldsCustom fields (project-level)
/api/v/favoritesFavorites toggle + order
/api/v/sectionsSections CRUD, reorder, task move
/api/v/homeHome page stats
/api/v/attachmentsFile upload + management (download only when authenticated)
/api/v/recurring-issuesRecurring tasks
/api/v/dataWorkspace JSON export / import, task import (CSV / Excel / Trello) and its undo
/api/v/audit-logsActivity log (incl. CSV export)
/api/v/api-tokensAPI token CRUD
/api/v/webhooksWebhook CRUD + delivery log

Example — List tasks

bash
curl -H "Authorization: Bearer $SPIRE_TOKEN" \ -H "Content-Type: application/json" \ "https://spire.hu/api/v/issues?status=in_progress&per_page=25"

Available query parameters: status, priority, type, assignee_id, project_id, cycle_id, parent_id, search, page, per_page (max 100).

json
{ "data": [ { "id": "01J7...", "identifier": "ACME-42", "title": "Payment flow breaks after 3D Secure", "status": "in_progress", "priority": "high", "type": "bug", "assigneeId": "01J7...", "dueDate": "2026-04-25", "createdAt": "2026-04-18T10:23:11Z", "updatedAt": "2026-04-19T09:11:42Z" } ], "meta": { "total": 1, "page": 1, "per_page": 25, "total_pages": 1 } }

Error codes

CodeMeaning
400Invalid request — see error.code
401Missing or expired token
403No permission to the given resource; LIMIT_REACHED — the workspace's AI credits or a monthly AI cap are used up
404The resource doesn't exist or isn't in your workspace
409Conflict (e.g. taken slug)
429Rate limit exceeded
500Server error — if it repeats, write to silner@spire.hu
503Temporarily unavailable (e.g. AI provider failure or the rate limiter being down)

API tokens and scopes

On the Settings → API tokens page you give a token a name, an access level (read only, or read + write) and an expiry. Owners and admins may also create an unrestricted token (optional expiry); members always create a limited token with an expiry. Every token reaches the /api/v/* endpoints with at most its creator's workspace permissions — a member's token cannot reach admin-only endpoints. At most 20 live tokens per person per workspace. A finer-grained token can be created with POST /api/v/api-tokens (from a signed-in session, not with an API token), passing a scopes list of <resource>:read or <resource>:write entries, where the resource is the path segment after /api/v/ (e.g. issues:read, projects:write) or *. read allows GET/HEAD/OPTIONS requests, write allows everything. The catalog's purchase price is a separate grant: it comes only with an explicit catalog-costs:read (catalog-costs:write to edit it); neither * nor an unrestricted token carries it.

These routes are never reachable with an API token, whatever its scopes: /auth, /api/profile, /api/gdpr, /api/workspaces, /api/invitations, /api/v/api-tokens and the internal administration endpoints.

Webhooks

Webhooks are currently managed through the API only (/api/v/webhooks, owner + admin): create, update, delete, test delivery and the delivery log. In production the destination must be a public https URL. The signing secret is returned on creation. Available events:

issue.created, issue.updated, issue.deleted, issue.status_changed, project.created, project.updated, cycle.started, cycle.completed, member.invited, member.removed, partner.created, partner.updated, contact.created, catalog_item.created, catalog_item.updated, document.viewed, document.accepted, document.declined (they carry only an id; fetch the details through the API).

Every delivery is a POST request with these headers: X-Spire-Signature (sha256= + the HMAC-SHA256 of the body with the secret), X-Spire-Event, X-Spire-Webhook-Id and X-Spire-Delivery. On transient failures (network, timeout, 5xx, 408, 429) there are at most 3 attempts, waiting 5 and 30 seconds in between. Deliveries come from a durable queue (a restart loses no event) and run in parallel, so the order is not guaranteed and an event can occasionally arrive twice — deduplicate on the X-Spire-Delivery header.

Recurring tasks

Recurring task templates can currently be created through the API only (/api/v/recurring-issues; not available to guests): daily, weekly, bi-weekly, monthly or quarterly, following the workspace time zone. New tasks are created in the name of the template's creator. Missed occurrences (e.g. during downtime) are not caught up — only the next due task is created. If the creator loses access (leaves the workspace, becomes a guest, or can no longer see the project), the template is deactivated automatically.

Rate limits

Limits apply per IP address and endpoint — except API-token requests, which share one budget per token across all endpoints. Sign-in has additional per-account protection: 10 failed attempts for the same email from one IP range within 15 minutes trigger a temporary lock, attempts from elsewhere get an increasing delay, and a browser that signed in successfully before is exempted from the delay by a "known device" cookie. Re-authentication allows 5, 2FA codes 5 failed attempts per 15 minutes.

EndpointLimit
POST /auth/register10 / hour, and 3 / hour per e-mail domain
POST /auth/login, POST /auth/2fa/verify10 / 15 min
POST /auth/google50 / 15 min
POST /auth/refresh60 / 15 min
POST /auth/forgot-password, /auth/reset-password5 / hour
POST /auth/verify-email20 / hour
POST /auth/resend-verification5 / hour
POST /api/profile/change-password, /password/set10 / hour
POST /api/profile/2fa/setup20 / hour
POST /api/profile/google/connect5 / 15 min
DELETE /api/profile/google10 / 15 min
POST /api/gdpr/delete-account5 / 15 min
POST /api/workspaces/:id/transfer-ownership5 / 15 min
POST /api/v/members/invite30 / hour
POST /api/v/members/bulk-invite5 / hour
/api/v/ai/*15 / min (+ monthly AI allowance)
/api/v/data/export/workspace, /api/v/data/import, /api/v/data/import/tasks5 / hour
/api/v/data/import/preview, /api/v/data/import/tasks/preview20 / hour
POST /api/v/data/import/batches/:id/undo10 / hour
POST /api/v/feedback10 / hour
every /api/* request with an API token (except MCP)120 / min / token
Interactive API reference: the full public API (parameters, response schemas, examples, try-it-out button) is available at /en/docs/api — rendered by Scalar, generated automatically from the OpenAPI 3.1 spec. The raw spec is downloadable at /docs/api/openapi.json (importable into Apidog, Postman, Insomnia, or any third-party code generator). For concrete help, write to silner@spire.hu.

AI assistant (MCP)

Connect your own AI assistant — claude.ai, ChatGPT, Claude Code, Cursor — to spire: it searches and summarizes issues and, if you allow it, creates or updates them.

spire offers an MCP endpoint (Model Context Protocol): https://spire.hu/api/mcp. The assistant connects with a separate, limited API token and sees exactly what you see — issues of a private project only if you are a member, as a guest only the projects shared with you.

  1. Create a token

    Settings → API → Connect an AI assistant (any member except guests; the assistant works with your permissions). By default the token is read-only (issues, projects, cycles, members, labels); with Allow writing it can also create and update issues and comment on your behalf. The token expires in 90 days by default (at most 365 days; a token without an expiry is not accepted by the AI endpoint).

  2. Add it to your assistant

    After creation you get the ready-made command / config next to the token:

bash
# Claude Code claude mcp add --transport http spire https://spire.hu/api/mcp \ --header "Authorization: Bearer spire_…"
json
// Cursor: .cursor/mcp.json — the token comes from the SPIRE_TOKEN environment variable { "mcpServers": { "spire": { "url": "https://spire.hu/api/mcp", "headers": { "Authorization": "Bearer ${env:SPIRE_TOKEN}" } } } }

claude.ai and ChatGPT (sign in, no token)

In claude.ai add a custom connector, in ChatGPT a connector in developer mode, with the URL https://spire.hu/api/mcp — no client id or secret is needed. The assistant opens the spire consent page: sign in, choose the workspace (any where you are not a guest — the assistant works with your permissions) and decide whether it may write (off by default). The app name on that page is given by the app itself — check the address it returns to, and only continue if you started the connection from the app yourself. The workspace owner gets a security e-mail about new connections. You see and can revoke your own connections at any time under Settings → Connected apps (owners and admins see every member's); the assistant loses access immediately. Claude Code works this way too (claude mcp add --transport http spire https://spire.hu/api/mcp without a header).

What the assistant reads goes to the external provider you chose, and your company is responsible for it. You acknowledge this on the consent page and when you create a token that reaches MCP (the page lists the modules whose data the connection reaches). The owner or an admin can switch external AI assistants off under Settings → AI usage: no new connection can be made, and existing ones cannot reach the workspace until it is switched back on.

Tools

AccessTools
Readwhoami, list_projects, get_project, search_issues, get_issue, list_my_issues, list_comments, list_cycles, get_cycle, list_members, list_labels
Write (only if allowed)create_issue (subtasks too), update_issue, add_comment — with the same rules as in the app (a guest may only change the status of an issue assigned to them); automations run as usual
Time trackingget_time_report, list_time_entries, get_estimate_vs_actual (read; only your own time — an admin sees the team's only with a separately ticked team-insights scope), and if you allowed writing: start_timer, stop_timer, log_time — they always record for you and never change an existing or already invoiced entry
Partner registerOnly when a module (e.g. CRM or invoicing) uses the partner register and you have access to it: search_partners, get_partner, search_contacts, get_contact, list_activities, validate_partner_identifiers (read), and if you allowed writing: create_partner, update_partner, create_contact, update_contact, log_activity. Deleting, anonymizing, exporting, bank accounts and consent stay in the app
CatalogOnly when a module uses the catalog and you have access to it: search_catalog, get_catalog_item (read, with list prices), and if you allowed writing, with sales or inventory: create_catalog_item, update_catalog_item. The assistant sees the purchase price only with a separately ticked catalog-costs scope and never writes it; deleting, deactivating and categories stay in the app
DocumentsRead-only, and only with the issuing module's access: get_document_status (an entity's document: versions, links, decision) and list_documents. The PDF content and the link token never reach the assistant; issuing, creating links and changing the settings stay in the app
Data the assistant reads is sent to the AI provider you chose (e.g. Anthropic, OpenAI), under their terms. Only connect a workspace whose data you may process there.
Prompt injection: anyone in the workspace can write issue, description and comment text. The assistant receives it as data (marked separately), but a malicious text may try to "instruct" it. Only grant write access if you approve every action of the assistant, and do not combine it with other sensitive tools.

Security and limits

Only a limited (scoped, expiring) API token or a connector sign-in works — not an unrestricted token and not a browser session. Revoking the token (Settings → API → Revoke) disconnects the assistant immediately; if its owner leaves the workspace or is locked, the token stops working too. Limits: 120 calls / minute per token, 200 / minute per user (all tokens together), writes 30 / minute per token. Actions appear in the audit log with the token id. A connector sign-in issues a 1-hour access token that only works on the MCP endpoint and is renewed automatically; removing the member who connected it, or making them a guest, ends the connection (and deletes their API tokens) at once.