Summary
Prioritisation, forecasting and PRDs that read from your actual backlog.
The nine tools chain naturally: query the backlog, score it against strategy, then forecast what that ordering means for a ship date. Every answer depends on what PM33 already knows about your product, so the quality tracks how complete your workspace is.
What it is
The MCP server for PM33, an AI product-management platform: nine tools over your backlog, velocity, strategy and competitive data, with nine resources and four interactive views alongside them.
What you get
- The backlog queried, scored with WSJF, and checked against strategy (pm33_query_backlog, pm33_optimize_priorities, pm33_score_alignment)
- A delivery forecast with confidence intervals, built from measured velocity (pm33_analyze_velocity, pm33_forecast_delivery)
- What-if scenarios on scope, timeline or resources, with delivery, cost and risk impact modelled (pm33_analyze_scenario)
- Epics scheduled across team lanes, respecting velocity and dependencies (pm33_schedule_portfolio)
- A PRD generated with strategic context and the competitive landscape (pm33_generate_prd)
- Competitive intelligence alerts with recommended actions (pm33_competitive_threats)
Requirements
A PM33 account. Either an API key generated at pm-33.io/settings and put in the environment, or OAuth with PKCE, which prompts you to log in on first use.
Setup effort
One command plus a key — claude /install-plugin https://github.com/b33-steve/pm33-mcp, then supply credentials