MCP

The Pome MCP — the Pome Control MCP, or “coach MCP” — is a hosted Model Context Protocol server at https://mcp.pome.sh/mcp. Transport is Streamable HTTP only (no SSE resumability). It lets any MCP client — Claude Code, Claude Desktop, claude.ai — author scenarios, launch evaluation runs, and read back scored reports, entirely through tool calls. This is what makes Pome usable by an AI agent directly, not just a human at a CLI.

Coach and examinee

The agent you connect to the MCP is the coach. It authenticates, picks or authors a scenario, launches a run, and reads the report. The coach can see grading criteria. The examinee is a faithful clone of the agent-under-test that Pome launches separately — for example, on Anthropic Managed Agents — with its service MCP URLs swapped to Pome digital twins. The examinee holds zero Pome tools and never sees the criteria. run_scenario hands the coach a launch spec for the clone; the coach launches it.

First call is always get_started. It returns the coach playbook.

Tool index

Orientation

Tool	What it does
`get_started`	Return the coach playbook. Call this first.
`get_platform_context`	Identity, team, plan tier, quota usage.
`list_twins`	The digital-twin catalog available on hosted.

Authoring

Tool	What it does
`list_scenarios`	Team scenario catalog with criteria.
`validate_task`	Grammar-only validation of scenario markdown.
`save_scenario`	Upsert a scenario.
`evaluate_criteria`	Dry-run `[code]` criteria against the seed without running the examinee.
`verify_seed`	Flag criteria that already pass on the seed — a broken seed.
`register_agent`	Lightweight registration of the agent under test.
`intake_clone_scope`	Register an agent, capture its clone scope, and mint a twin façade token.
`list_agents`	List registered agents.

Run lifecycle

Tool	What it does
`run_scenario`	Provision one twin sandbox per twin, return the examinee launch spec.
`run_trials`	Batch: N trials of one scenario sharing a group.
`list_sessions`	List recent twin sessions.
`get_session`	Side-effect-free session probe.
`stop_session`	Stop a session early without evaluating.
`finalize_run`	Grade synchronously; returns the score directly — no polling.
`check_evaluation`	Poll a legacy async evaluation job.
`get_report`	Scored markdown report and an app.pome.sh weblink.
`list_runs`	List runs newest-first.

Connect

Add the server and authenticate.

Tools

Full reference for every tool.

Workflow

The end-to-end coach workflow.

Digital Twins reference

The deterministic, resettable SaaS services run_scenario provisions against.

Get started

Guides

Skills

Command Line Interface

MCP

Coach and examinee

Tool index

Orientation

Authoring

Run lifecycle

Connect

Tools

Workflow

Digital Twins reference

​Coach and examinee

​Tool index

​Orientation

​Authoring

​Run lifecycle

Connect

Tools

Workflow

Digital Twins reference

Coach and examinee

Tool index

Orientation

Authoring

Run lifecycle