@topy-ai/cpo-mcp

TOPY.AI's AI CPO as a local stdio MCP server, for MCP clients that only run local servers. Every message goes to https://topy.ai/mcp/cpo with your TOPY sign-in. Clients with remote MCP can use that URL directly.

{ "mcpServers": { "topy-cpo": { "command": "npx", "args": ["-y", "@topy-ai/cpo-mcp"] } } }

Sign in once with npx @topy-ai/cpo-mcp login (shared with cpo-cli, cfo-cli and ceo-cli), or set TOPY_API_KEY (Dashboard → API keys).

Tools

• Product (reads): cpo_review (the weekly product review), cpo_status_report (KPIs, risks and asks for the CEO), cpo_feedback_list, cpo_decisions, cpo_experiments, cpo_prioritize (RICE), cpo_experiment_size (sample size and duration), cpo_pmf_plan (a product–market-fit check), cpo_positioning.
• Changes (preview first): cpo_feedback_add, cpo_feedback_triage (accept, reject, duplicate, needs info; size; what it touches), cpo_decision_log (every decision with a falsifier), cpo_experiment_start (pre-registered: hypothesis, metric, MDE, sample, stop rule), cpo_experiment_read (no peeking: before the planned sample it says "keep running"), cpo_spec (a PRD with a success metric and how it is measured), cpo_roadmap (now, next, later), cpo_build_check (build, buy or skip), cpo_report_send (the weekly review to the Founder or the CEO).

Writes take two calls: a preview, then the same call with confirm: true after you agree. Usage numbers and results come from the tools, never invented. The CPO decides what to build; it writes no code — the CTO builds. Prompts (workflows) cover the weekly review, feedback triage, an experiment, a PRD and build-or-not.

MIT · topy.ai/developers