Bring your own agent

​

Start with the example

​

Before you start

• The quickstart’s setup, done once: the coach skills installed (npx skills add pome-sh/digital-twins) and the Pome MCP connected (claude mcp add --transport http pome https://mcp.pome.sh/mcp).
• The Pome CLI, which writes and reads the manifest:
  npm install -g @pome-sh/cli
  pome login
  
  pome login opens app.pome.sh, signs you in, and stores a pme_… key in your macOS Keychain (or ~/.pome/credentials.json). CI can set POME_API_KEY instead.
• Your agent’s repo, open in Claude Code — the one you want to test. It should talk to a service Pome has a twin for (GitHub, Slack, Stripe, Gmail, Linear), and you should know how it starts (e.g. npm start).

​

Register your agent

Connect this repo to Pome as my own agent — take it up to a registered agent
with a written pome.json, and stop there. Do not run it yet.

1. Confirm I've already seen a scored run on the bundled support-triage
   example (the quickstart). If I haven't, tell me to do that first — the
   example is how the loop is taught.
2. Scaffold the manifest so I never hand-write it: run `pome init` in this
   repo (use `pome init --sdk claude` if this is a Claude Agent SDK project).
   That writes a starter pome.json.
3. Set `command` in pome.json to how THIS agent actually starts — detect it
   from package.json / the repo (e.g. `npm start`) and replace the scaffold
   default. Confirm the command with me before writing it.
4. Register the agent so runs scope to this project:
   `pome register agent "<a short name>" --twins <services it uses, e.g. github,slack>`.
   This POSTs the identity and writes the server-canonical agent.slug back
   into pome.json.
5. Show me the final pome.json (agent.slug + command + twins) and the
   dashboard line register printed. Stop here — the first run is the next step.

Guardrails: keep every command plain text and show it to me before running;
never write secrets or URLs into source. pome.json is safe to commit; the
agt_ id register caches under .pome/ is not — leave it gitignored.

​

What you have now

• pome.json — agent.slug, the real start command, and the twins this agent may exercise. This is the manifest every run reads.
• A registered agent on app.pome.sh, scoped to your team.

1. Preflight the wiring. Run pome doctor in the repo. It runs four checks in order — valid pome.json command, twin reachable, no production host surviving in agent source, egress locked to the twin — and stops at the first red with one named cause and one fix. A green preflight is the gate: don’t launch a run until pome doctor exits clean. See Troubleshooting.
2. Your first run. With the wiring green, the coach suggests a first task matched to your agent and takes it to a scored report — the same run → read the failure → fix → re-run loop the example taught you. The MCP workflow is that loop, call by call.

​

Next