PM33 MCP Server

Connect Claude to your product management workflow — backlog optimization, delivery forecasting, competitive intelligence, and PRD generation.

Quick Start

Claude Desktop (OAuth — recommended)

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "pm33": {
      "url": "https://pm-33.io/mcp",
      "transport": "streamable-http"
    }
  }
}

Claude Desktop auto-discovers OAuth, opens a browser for consent, and caches tokens. No API key needed.

Claude Code (CLI)

claude mcp add pm33 --transport streamable-http https://pm-33.io/mcp

API Key (programmatic access)

{
  "mcpServers": {
    "pm33": {
      "url": "https://pm-33.io/mcp",
      "transport": "streamable-http",
      "headers": {
        "x-api-key": "pm33_live_xxxxxxxxxxxx"
      }
    }
  }
}

Generate an API key from Settings > Integrations > MCP Server in PM33.

Tools

PM33 provides 9 tools designed for multi-tool chaining. Claude reads these descriptions to decide which tools to invoke and in what order.

query_backlog read

Retrieve work items from the team's backlog. Start here to get item IDs needed by optimize_priorities, score_alignment, or schedule_portfolio. Supports active/full mode with filtering, pagination, and configurable column sets.

Start here → then chain with any scoring or scheduling tool

optimize_priorities read

Answer "What should we build next?" — Scores backlog items using research-based WSJF with strategic alignment, cost of delay, and competitive threat factors. Returns P0–P4 tiers.

Use after query_backlog → then chain with schedule_portfolio

schedule_portfolio read

Schedule epics across team lanes using PM33's TieredScheduler. Allocates work respecting velocity, skills, and dependencies. Enable includeStoryPacking for sprint-level assignments.

Use after optimize_priorities to schedule the prioritized stack

forecast_delivery read

Answer "When will X ship?" — Runs PM33's multimodal optimization engine for probabilistic delivery date forecasts with confidence intervals (P50, P75, P90, P95).

Use analyze_velocity first to establish capacity context

analyze_velocity read

Understand team delivery capacity and trends. Returns velocity, predictability score, and bottleneck analysis.

Use before forecast_delivery to ground forecasts in actual team data

competitive_threats read

Answer "What are competitors doing?" — Surfaces competitive intelligence alerts with recommended actions. Tenant-scoped (no workspaceId needed).

Use before optimize_priorities with includeCompetitiveThreats=true

score_alignment AI credits

Evaluate how well work items align with strategic objectives. Returns alignment level, confidence, and reasoning per item. Consumes AI credits.

Use after query_backlog to score specific items

generate_prd AI credits

Answer "Write the PRD for this feature" — Generates a PRD using backlog context, strategic objectives, and competitive landscape. Set includeStrategicContext=true for full enrichment.

analyze_scenario read

Answer "What if we add/cut/delay X?" — Models the impact of resource, scope, or timeline changes against the current plan. Returns delivery date shifts, cost impact, and risk score changes.

Resources

5 data resources provide read access to workspace data, plus 4 interactive UI views.

URI	Description	Cache TTL
`pm33://strategy/{workspaceId}`	Strategic context — mission, objectives, themes	300s
`pm33://backlog/{workspaceId}/summary`	Backlog summary — item counts, status breakdown	60s
`pm33://velocity/{workspaceId}/current`	Current velocity snapshot — throughput, predictability	120s
`pm33://competitive/{tenantId}/landscape`	Competitive landscape — tenant-scoped competitor data	300s
`pm33://workspaces`	Available workspaces for the authenticated user	30s

Interactive Views (MCP Apps)

4 self-contained HTML views render inline in Claude Desktop, Claude.ai, and VS Code. No external dependencies.

View	URI	Visualization
Portfolio Schedule	`pm33://ui/schedule-portfolio`	CSS Grid Gantt chart with team lanes and utilization heatmap
Delivery Forecast	`pm33://ui/forecast-delivery`	Chart.js bar + doughnut for confidence intervals and risk
Priority Dashboard	`pm33://ui/optimize-priorities`	Chart.js distribution + SVG radar + interactive scored items table
PRD Document	`pm33://ui/generate-prd`	Markdown renderer with TOC sidebar and strategic context

Authentication

OAuth 2.0 (recommended)

PM33 supports OAuth 2.0 Authorization Code with PKCE (S256 only). MCP clients auto-discover the OAuth configuration via standard endpoints:

Endpoint	URL
Protected Resource Metadata (RFC 9728)	`/.well-known/oauth-protected-resource`
Authorization Server Metadata (RFC 8414)	`/.well-known/oauth-authorization-server`
JWKS	`/.well-known/jwks.json`
Authorization	`/oauth/authorize`
Token	`/oauth/token`
Revocation	`/oauth/revoke`

Token lifetime: Access tokens expire after 15 minutes. Refresh tokens last 30 days and rotate on each use.

API Key (alternative)

For programmatic access, pass an API key in the x-api-key header. Generate keys from Settings > Integrations > MCP Server. Keys must have at least the mcp:enabled scope.

Scopes

Scope	Grants Access To
`mcp:enabled`	Base MCP access (required)
`backlog:read`	Backlog items, epics, work item queries
`optimize:read`	WSJF scoring, backlog optimization
`velocity:read`	Sprint velocity, forecasting, Monte Carlo simulations
`competitive:read`	Competitive intelligence analysis
`strategy:read`	Strategic alignment scoring
`scheduling:read`	Portfolio scheduling, lane allocation
`scenarios:read`	What-if scenario modeling
`prd:write`	AI-powered PRD generation (consumes AI credits)

Example Workflows

Ask Claude natural questions — it chains the right tools automatically:

"What should we build next?"

query_backlog → optimize_priorities → returns ranked P0–P4 items with interactive priority dashboard

"When will the auth redesign ship?"

analyze_velocity → forecast_delivery → returns P50/P75/P90/P95 dates with forecast chart

"Write a PRD for real-time collaboration"

generate_prd with includeStrategicContext=true → returns rendered PRD with strategic alignment

"What if we add 2 engineers to the platform team?"

analyze_scenario → returns delivery date shifts, cost impact, and risk changes

"Schedule our prioritized backlog across teams"

query_backlog → optimize_priorities → schedule_portfolio → returns Gantt chart with lane assignments

Troubleshooting

Issue	Resolution
Connection refused	Verify `https://pm-33.io/mcp` is accessible
OAuth consent doesn't appear	Check browser pop-up blocker; ensure PM33 account exists
"Invalid scope"	Ensure `mcp:enabled` is included in requested scopes
"Tenant not found"	User must be a member of at least one PM33 workspace
Token expired	Client auto-refreshes; if persistent, re-authorize
API key rejected	Ensure key has `mcp:enabled` scope and is not revoked
AI tool returns "insufficient credits"	Top up workspace AI credits in Settings > Billing

PM33 MCP Server v1.2.0 — 9 tools, 9 resources, 4 interactive views
Privacy Policy · Terms of Service · support@pm-33.io