Spedy Docs

Integration layer to trigger your Anthropic Claude Routines directly from a Spedy ticket — with ticket-context variables, optional board/repo scope, pre- and post-actions, and results posted back into the ticket.

Claude Routines are an Anthropic product: reusable work instructions you define in your Anthropic account and Claude executes for you. Spedy doesn't host or run the routines — Spedy provides the integration layer so you can trigger one of your routines directly from a ticket.

Concretely, this integration:

resolves ticket context (title, description, status, board, URL) into your prompt template,
POSTs the resolved prompt to your Anthropic routine webhook with your secret,
runs optional pre- and post-steps on the ticket (status change, self-assign, PR-link comment),
tracks each execution and surfaces its status on the ticket.

The routines themselves, the model usage, and the cost remain in your Anthropic account.

The Spedy integration is part of the Pro plan and must be enabled at the organization level via the CLAUDE_ROUTINES feature flag.

Concepts

In Spedy, a routine entry has four layers:

Metadata — name, slug, description
Scope — optional board and repository scope
Instructions — the prompt template (supports variables like {{ticket.title}})
Pre- and post-actions — server-side ticket steps that run before/after the hand-off to Claude

Every trigger creates a RoutineExecution that tracks the status of the hand-off, the resolved prompt, and completion metadata. The actual AI work happens on Anthropic — RoutineExecution is only Spedy's view of it.

Permissions and Feature Flag

Routines are gated by:

Gate	Value	Scope
Organization feature flag	`CLAUDE_ROUTINES`	All `/routines/*` endpoints and UI require this to be enabled
Permission `routines:manage`	Default for Admin + Team Member	Create/read/update/delete your own routines
Permission `routines:trigger`	Default for Admin + Team Member	Trigger and cancel executions on tickets you can access

Customers (organization role CUSTOMER) are explicitly excluded from both permissions.

Board-level access is also enforced: triggering a routine on a ticket whose board you cannot access returns 403 Forbidden.

Prompt Variables

Inside instructions, the following {{variable}} placeholders are resolved at trigger time:

Variable	Source
`{{ticket.id}}`	Ticket ID
`{{ticket.displayId}}`	Human-readable ticket key (e.g. `WEB-42`)
`{{ticket.title}}`	Ticket title
`{{ticket.description}}`	Ticket description
`{{ticket.status}}`	Status name
`{{ticket.statusKey}}`	Status key (`BACKLOG` … `DONE`)
`{{ticket.boardId}}`	Board ID
`{{ticket.boardName}}`	Board name
`{{ticket.boardSlug}}`	Board prefix
`{{ticket.url}}`	Relative URL to the ticket

If the trigger request includes an optional text field, it is appended to the resolved prompt under a --- separator as Additional instructions for this run:.

Pre- and Post-Actions

Three server-side action types are supported in preActions and postActions:

Type	Params	When it runs	Effect
`move_status`	`statusKey`: `BACKLOG` \| `TODO` \| `IN_PROGRESS` \| `BLOCKED` \| `DONE`	pre / post	Transitions the ticket to the matching board status
`assign_self`	—	pre / post	Assigns the triggering user as the ticket assignee
`post_pr_comment`	—	post	Posts the returned PR link as an internal comment on the ticket

Actions execute in order. Pre-actions run synchronously before Claude is notified; post-actions run when Claude reports completion.

Scoping

Two fields narrow where a routine may run:

boardIds — if non-empty, the routine is only offered on tickets whose board is in this list. A trigger against a board outside the list returns 400.
allowedRepos — a list of owner/repo strings. Only repositories that are linked to the ticket's board and appear in this list are passed to Claude as repo context. An empty list means "no repo constraint".

Use the GET /routines/available-repos helper to discover valid allowedRepos values for your organization.

Outgoing Webhook

To hand off the work to Anthropic's Claude Code Routines endpoint, each routine can set:

Field	Description
`webhookTargetUrl`	HTTPS URL on `api.anthropic.com`. On trigger, Spedy POSTs `{ "text": resolvedPrompt }` to this URL.
`webhookTargetSecret`	Sent as `Authorization: Bearer <secret>`. Never returned in list/read responses (only the boolean flag `webhookTargetSecretSet`).
`webhookTargetHeaders`	Additional non-secret headers merged into the outgoing request (e.g. `{"anthropic-version": "2023-06-01"}`). Spedy supplies sensible defaults for `api.anthropic.com` if you don't.

On a successful hand-off the execution flips from PENDING to CLAIMED (with claimedAt set), and the ticket card shows Handed off to Claude.

API Reference

All endpoints are mounted under /api/v1/routines, protected by JwtAuthGuard, and support Personal Access Tokens.