Every tool the MCP server exposes, grouped into read tools — for pulling context, finding work, and checking status — and create tools — for opening tickets and projects — and update tools for commenting and changing status. Connect a tool first (Setup & usage), then call these from your editor.
01 How the tools work
Every tool is namespaced leroy_ so it never collides with the other tools in your client. Each one is scoped to the account you signed in as— it only ever sees that account's tickets, projects, and PRs. To work across accounts, add the server again and sign in to the other one.
The heavier read tools accept a response_format argument:
Format
When to use
CONCISE
Default. A trimmed view — enough to answer a question or pick a ticket without flooding the context window.
DETAILED
The full payload — every field, the whole conversation, all references. Use it when you're about to start coding.
i
Tip.You don't call these tools by hand — your AI tool picks them. Ask in plain English ("use leroy to find the auth tickets") and it chooses the right one and arguments.
02 Read tools
Pull context, find work, and check what's in flight.
Tool
What it does
leroy_ask_leroy
Ask a natural-language question about your pipeline; get a synthesized answer plus the referenced ticket ids.
leroy_find_tickets
Structured filter over tickets — by status, keyword, or ref.
leroy_search
Keyword / full-text search across ticket titles, bodies, and conversations.
leroy_get_ticket
Full detail for one ticket: status, body, conversation, plan, linked branch and PR.
leroy_get_ticket_context
A context pack to cold-start a ticket locally: body, related code references, prior decisions, and a suggested base branch.
leroy_get_plan
A ticket's execution plan: per-repo branches and PRs, what's done vs. left, and any blocking questions.
leroy_get_execution
A ticket's agent run(s): current stage, status, what's blocking, branch and PR, and cost.
leroy_my_attention
What Leroy is waiting on you for: PR reviews, plan approvals, and clarifications.
leroy_list_pull_requests
Pull requests Leroy has opened across the account.
leroy_list_projects
List the account's projects.
leroy_get_project
Full detail for one project: docs, tickets, and lifecycle.
03 Create tools
Open new work from your editor.
Tool
What it does
leroy_create_project
Create a project, optionally seeded with spec docs and tickets.
leroy_create_ticket
Create a ticket: agent-owned (Leroy works it now), agent-owned but not started, human-owned, or parked.
When you create an agent-owned ticket, it still passes the policy gates for the account — Leroy may pick it up immediately or wait for your go-ahead depending on how intake is set. See Policy gates.
prompt
Use leroy to open a ticket: fix the flaky checkout test in payments-api. Don't start it yet.
04 Update tools
Comment on a ticket or move its status — without leaving your editor.
Tool
What it does
leroy_comment_on_ticket
Post a comment onto a ticket's conversation. Unless the ticket is parked, this re-engages Leroy — it reads the comment and responds or acts, so a reply can answer a clarification or steer in-flight work.
leroy_update_ticket_status
Set a ticket's status: not started, in progress, blocked, parked, done, or cancelled. A linked Linear issue moves to its matching state.
i
Which states you can set.leroy_update_ticket_status only sets the states you manage. The agent-pipeline states (coding, reviewing) and the review gates (approve to start, review plan, in review) are driven by the work itself — to hand a ticket to Leroy, create it agent-owned or comment to re-engage rather than forcing a status.
prompt
Use leroy to comment on PIT-693: the retry budget should be 3, not 5 — please update the plan.
!
One account per connection. The MCP is scoped to the account you signed in as. To work across accounts, add the server again and sign in to the other one.