Claude Routines on Your Tickets: One Click, Claude Takes Over

Note (2026-09-02): Claude Routines have been removed from Spedy. Spedy no longer triggers agents — your own agent (Claude Code, Codex, Cursor, …) works with Spedy directly via MCP. Background in Runners, live previews and agent skills removed. The text below is kept as the historical announcement.

Many of you already work with Claude Routines — Anthropic's feature for reusable, Claude-driven work instructions. The sore spot until now: you see a ticket in Spedy, want to run exactly that routine on it, and still end up copy-pasting context into Claude, firing the routine by hand, and manually writing up what came out.

That bridge now exists: trigger Claude Routines straight from a Spedy ticket. Wire up the routine once, then a single click is enough. Spedy sends Claude the fully resolved prompt with ticket context, tracks the execution, and posts the PR link back into the ticket.

Up front, so there's no confusion: the routines themselves are and remain your Anthropic feature. Spedy doesn't build them — Spedy integrates them.

The integration has to be enabled by an org admin via the CLAUDE_ROUTINES feature flag.

---

What You Can Do Now

• Set up once, use everywhere: for each of your routines you create an entry in Spedy — name, prompt template with ticket variables, your Anthropic webhook URL, your Anthropic secret.
• Trigger with one click: every ticket gets a Claude Routines card. Pick a routine, optionally add extra text, hit Trigger — done.
• Track status on the ticket: running and past executions show up right on the ticket. You can cancel your own runs.
• Automatic feedback: when Claude is done, the PR link lands as an internal comment on the ticket — so you instantly see the outcome without switching tabs.

---

What Spedy Handles for You

So you don't have to copy-paste, the integration takes care of three things:

1. Ticket context in the prompt

Your prompt template can reference variables — Spedy fills them in before the request leaves:

Work on ticket {{ticket.displayId}}: {{ticket.title}}

{{ticket.description}}

Link: {{ticket.url}}

Available variables: {{ticket.id}}, {{ticket.displayId}}, {{ticket.title}}, {{ticket.description}}, {{ticket.status}}, {{ticket.statusKey}}, {{ticket.boardId}}, {{ticket.boardName}}, {{ticket.boardSlug}}, {{ticket.url}}.

On trigger you can also pass free-text ("also please write unit tests"). It's appended under a --- separator to the resolved prompt — just for that one run.

2. Pre- and post-steps on the ticket

A few things shouldn't be Claude's problem — so Spedy takes care of them around the trigger:

Step	When	What
move_status	before or after	Move the ticket to another board status (IN_PROGRESS, DONE, …)
assign_self	before or after	Assign the triggering user
post_pr_comment	after	Post the PR link as an internal comment on the ticket

Example: "before the run, move to IN_PROGRESS and assign me; after the run, drop the PR link as an internal comment." Configure it once and it happens on every execution automatically.

3. Scope: where a routine even shows up

Not every routine belongs on every board. That's why each routine has two optional limits:

• Boards — empty means "available on every board you can access". Non-empty means "the routine only shows up on tickets of these boards".
• Repositories — a list of owner/repo strings. Claude only receives repositories as context that are both linked to the ticket's board and listed here.

---

One-Time Setup (per User)

Under Account → Claude Routines, a 4-step wizard walks you through onboarding:

1. Create a Personal Access Token — so the MCP server and your routine runners can read/write on your behalf.
2. Drop in your webhook URL and secret — the URL Anthropic gave you for your routine. Spedy POSTs to exactly this URL on trigger.
3. Copy the CLAUDE.md block — a short snippet for your project repo so Claude knows the tickets_* MCP tools (read ticket, post comment, set status).
4. First run — trigger a simple routine on a test ticket, verify that the PR link comment appears.

After that it's one click per ticket.

---

Who Can Do What

The integration ships with two new permissions:

• routines:manage — create, edit, delete your own routines in Spedy.
• routines:trigger — run a routine on a ticket and see past executions.

Both permissions are in the default Admin and Team Member roles. Customers are intentionally excluded — they see neither the ticket card nor the account area.

There's also an org-level feature flag CLAUDE_ROUTINES. As long as that's off, the endpoints return 403 and the UI hides everything — an admin can flip it under org features.

And on every trigger Spedy does a last check: you can only run a routine on a ticket whose board you actually have access to.

---

Get Started

1. Admin: enable CLAUDE_ROUTINES in organization settings.
2. In Anthropic, set up a routine and keep the webhook URL + secret handy.
3. In Spedy, open Account → Claude Routines and run through the setup wizard.
4. Create your first routine in Spedy — with your prompt template, board scope, and repo scope.
5. On a matching ticket, open the Claude Routines card and hit Trigger.