Connect your terminal AI to Workcamp

A shared work-tracking layer for your AI coding tools. Once connected, your terminal AI can't just silently edit code — it works the way your team does in Workcamp Manager:

• Opens a ticket before it codes — every task becomes a Workcamp ticket, set In Progress and assigned to you.
• Logs its progress & decisions as comments on that ticket while it works.
• Records time spent — hours per date, just like your team logs time on tickets.
• Closes the ticket with a summary when the work is done and verified.
• Everything under your identity, in your workspace — managers see it all in Manager → AI Activity.

You stay in control: it asks before creating a project (never duplicates), and every action is scoped to the permissions you grant at sign-in.

Step 1 · Your operating system

Step 2 · Your AI harness

Not listed? Pick Other / generic MCP — the server is a standard remote MCP endpoint and any spec-compliant client can connect.

Step 3 ·

Setting this up with an AI?

Point it at this page and it will do the whole thing itself — detect your OS and client, run the installer, write the right config, and verify the connection.

Read https://mcp.workcamp.app/setup and connect this repo to Workcamp

Plain markdown, built to be read by an agent: mcp.workcamp.app/setup

Per repo — grab the governance templates

1. append CLAUDE.md (or AGENTS.md for Codex/others) to your repo's file
2. save .workcamp.yml at the repo root (the agent fills in projectId on first run)
3. optional: /workcamp → .claude/commands/workcamp.md

1. Before writing code, it opens a ticket (start_task) and sets it In Progress.
2. If there's no .workcamp.yml / known project, it asks you — link an existing project (it says who to request access from if needed), create a new one, or don't track this — never auto-creating duplicates. Your answer is pinned so it won't ask again.
3. Personal or planning work? Set tracking: off in a .workcamp.yml (or a per-user .workcamp.local.yml) and that folder is left untracked — no tickets, no time. Git isn't required; the pin works at a non-git or multi-repo root too.
4. It logs steps (log_progress), records time spent (log_time — hours per date, many entries per ticket), and closes the ticket (complete_task) when done.

Managers see all of it in Manager → AI Activity. The authoritative, always-current rules are whatever get_policy returns.

The rules update themselves. Behavior comes from the live policy (get_policy), so every session already runs the latest — no action needed when we ship a change.

The only local copies are the governance text in your repo (CLAUDE.md / AGENTS.md / /workcamp). The agent tells you when they're behind; refresh them in place any time (leaves your .workcamp.yml pins untouched):

macOS / Linux

curl -fsSL https://mcp.workcamp.app/install.sh | sh -s -- --update

Windows

$env:WORKCAMP_UPDATE=1; irm https://mcp.workcamp.app/install.ps1 | iex