ProductPlan MCP Server

1. Find your config file:

   • Mac: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\Claude\claude_desktop_config.json

2. Open it in any text editor and add this (replace your-token with your actual API token):

Mac/Linux:

{
  "mcpServers": {
    "productplan": {
      "command": "/usr/local/bin/productplan",
      "env": {
        "PRODUCTPLAN_API_TOKEN": "your-token"
      }
    }
  }
}

Windows:

{
  "mcpServers": {
    "productplan": {
      "command": "C:\\Tools\\productplan.exe",
      "env": {
        "PRODUCTPLAN_API_TOKEN": "your-token"
      }
    }
  }
}

3. Restart Claude Desktop

Add to your config file:

• Mac/Linux: ~/.claude.json
• Windows: %USERPROFILE%\.claude.json

Mac/Linux:

{
  "mcpServers": {
    "productplan": {
      "command": "/usr/local/bin/productplan",
      "env": {
        "PRODUCTPLAN_API_TOKEN": "your-token"
      }
    }
  }
}

Windows:

{
  "mcpServers": {
    "productplan": {
      "command": "C:\\Tools\\productplan.exe",
      "env": {
        "PRODUCTPLAN_API_TOKEN": "your-token"
      }
    }
  }
}

1. Open Cursor
2. Go to Settings → MCP Servers
3. Add this configuration:

Mac/Linux:

{
  "productplan": {
    "command": "/usr/local/bin/productplan",
    "env": {
      "PRODUCTPLAN_API_TOKEN": "your-token"
    }
  }
}

Windows:

{
  "productplan": {
    "command": "C:\\Tools\\productplan.exe",
    "env": {
      "PRODUCTPLAN_API_TOKEN": "your-token"
    }
  }
}

Windows users: Use double backslashes (\\) in the path. This is required because backslash is an escape character in JSON.

1. Install the Cline extension
2. Open VS Code settings (JSON) and add:

Mac/Linux:

{
  "cline.mcpServers": {
    "productplan": {
      "command": "/usr/local/bin/productplan",
      "env": {
        "PRODUCTPLAN_API_TOKEN": "your-token"
      }
    }
  }
}

Windows:

{
  "cline.mcpServers": {
    "productplan": {
      "command": "C:\\Tools\\productplan.exe",
      "env": {
        "PRODUCTPLAN_API_TOKEN": "your-token"
      }
    }
  }
}

1. Install the Continue extension
2. Add to your config file:
   • Mac/Linux: ~/.continue/config.json
   • Windows: %USERPROFILE%\.continue\config.json

Mac/Linux:

{
  "mcpServers": [
    {
      "name": "productplan",
      "command": "/usr/local/bin/productplan",
      "env": {
        "PRODUCTPLAN_API_TOKEN": "your-token"
      }
    }
  ]
}

Windows:

{
  "mcpServers": [
    {
      "name": "productplan",
      "command": "C:\\Tools\\productplan.exe",
      "env": {
        "PRODUCTPLAN_API_TOKEN": "your-token"
      }
    }
  ]
}

1. Set environment variable on your n8n instance:

   N8N_COMMUNITY_PACKAGES_ALLOW_TOOL_USAGE=true
   

2. Add an MCP Client node to your workflow
3. Configure:
   • Command:
     • Mac/Linux: /usr/local/bin/productplan
     • Windows: C:\Tools\productplan.exe
   • Environment Variables: PRODUCTPLAN_API_TOKEN=your-token
4. Connect to an AI Agent node

Example workflow: Slack Trigger → AI Agent (with MCP Client) → Slack Response

productplan-mcp-server/
├── cmd/productplan/main.go      # Entry point: token check, then MCP server or CLI
├── internal/
│   ├── api/                     # ProductPlan API client
│   │   ├── client.go            # HTTP client: auth, rate limiting, error wrapping
│   │   ├── transport.go         # Tuned HTTP transport
│   │   ├── cache.go             # In-process TTL read cache (singleflight, write invalidation)
│   │   ├── list.go              # Paged collection GETs and Ransack query encoding
│   │   ├── ids.go               # Typed resource IDs (BarID, RoadmapID, ...), each validated into a path segment
│   │   ├── path.go              # Routes and request paths built only from typed IDs
│   │   ├── safeseg.go           # Path-segment validation for user-supplied IDs
│   │   ├── endpoints*.go        # Endpoint methods (roadmaps, bars, ideas, launches, OKRs)
│   │   ├── bars_read.go         # Roadmap bars with lane enrichment and client-side filters
│   │   ├── bar_schema.go        # Roadmap legends/lanes/custom fields for bar writes
│   │   └── formatters.go        # Response projection for AI
│   ├── mcp/                     # MCP wiring over the official go-sdk
│   │   ├── sdk_server.go        # Serves the registry via go-sdk (stdio)
│   │   ├── sdk.go               # Converts local Tool -> SDK tool
│   │   ├── handler.go           # Registry: dispatch and panic recovery
│   │   ├── argkeys.go           # Rejects undeclared argument keys (did-you-mean)
│   │   ├── editdistance.go      # Levenshtein distance for suggestions
│   │   └── types.go             # Tool-authoring types
│   ├── tools/                   # Tool definitions and handlers
│   │   ├── registry.go          # Tool registration (name -> handler)
│   │   ├── definitions*.go      # Tool schemas and descriptions
│   │   ├── helpers.go           # typedHandler, manage-action dispatch
│   │   ├── filters.go           # List-tool filters -> Ransack predicates
│   │   ├── formatter.go         # List/item/action response summaries
│   │   ├── item_type.go         # Item nouns for summaries
│   │   ├── bar_planner.go       # Validates bar writes against the roadmap
│   │   ├── bar_names.go         # Legend/lane/custom field name resolution
│   │   ├── bulk_bars.go         # bulk_update/create/delete_bars
│   │   ├── roadmaps.go, bars.go, ideas.go, objectives.go, launches.go, utility.go  # handlers
│   │   └── types*.go            # Typed argument structs for handlers
│   ├── cli/                     # CLI commands (status, roadmaps, etc.)
│   │   └── cli.go
│   └── logging/                 # slog JSON handler setup (ts/level/msg)
│       └── logger.go
├── pkg/productplan/             # Reusable utilities
│   ├── retry.go                 # Exponential backoff with jitter
│   ├── ratelimit.go             # Adaptive rate limiting
│   ├── batch.go                 # Batched operations
│   ├── health.go                # Health reporting
│   ├── requestid.go             # Request tracing
│   ├── validation.go            # ID validation (Field.RequireID)
│   └── errors.go                # APIError and error suggestions
└── evals/                       # LLM evaluation test suite
    ├── runner.go, types.go
    ├── tool_selection.json
    ├── confusion_pairs.json
    └── argument_correctness.json

Requires Go 1.26 or newer.

git clone https://github.com/olgasafonova/productplan-mcp-server.git
cd productplan-mcp-server
go build -o productplan ./cmd/productplan

Build for all platforms:

# macOS Apple Silicon
GOOS=darwin GOARCH=arm64 go build -o dist/productplan-darwin-arm64 ./cmd/productplan

# macOS Intel
GOOS=darwin GOARCH=amd64 go build -o dist/productplan-darwin-amd64 ./cmd/productplan

# Linux
GOOS=linux GOARCH=amd64 go build -o dist/productplan-linux-amd64 ./cmd/productplan

# Windows
GOOS=windows GOARCH=amd64 go build -o dist/productplan-windows-amd64.exe ./cmd/productplan

Run all tests:

go test ./...

Run with coverage:

go test ./... -cover

Run benchmarks:

go test ./internal/... -bench=. -benchmem

Run evaluation suite:

./scripts/run-evals.sh

Coverage (measured 24-09-2026, go test ./... -cover):

Every production file scores 10.0 in CodeScene Code Health (two type-only files can't be scored).

50 tools available: 35 READ tools and 15 WRITE tools (12 action-based manage_* plus 3 bulk_* bar tools):

Read tools:

• Roadmaps: list_roadmaps, get_roadmap, get_roadmap_bars, get_roadmap_lanes, get_roadmap_milestones, get_roadmap_legends, get_roadmap_comments, get_roadmap_complete
• Bars: get_bar, get_bar_children, get_bar_comments, get_bar_connections, get_bar_links
• OKRs: list_objectives, get_objective, list_key_results, get_key_result
• Discovery: list_ideas, get_idea, list_all_customers, list_all_tags, list_opportunities, get_opportunity, list_idea_forms, get_idea_form
• Launches: list_launches, get_launch, get_launch_sections, get_launch_section, get_launch_tasks, get_launch_task
• Admin: check_status, health_check, list_users, list_teams

Write tools:

• Roadmaps: manage_bar, manage_lane, manage_milestone
• Bar relationships: manage_bar_connection, manage_bar_link
• Bulk bars: bulk_update_bars, bulk_create_bars, bulk_delete_bars (up to 100 bars per call, validated up front, dry_run supported)
• OKRs: manage_objective, manage_key_result
• Discovery: manage_idea, manage_opportunity
• Launches: manage_launch, manage_launch_section, manage_launch_task

Example:

{"tool": "list_roadmaps", "arguments": {}}
{"tool": "manage_bar", "arguments": {"action": "create", "roadmap_id": "123", "lane": "Backend", "name": "New feature", "legend": "Committed"}}
{"tool": "bulk_update_bars", "arguments": {"set": {"legend": "Committed"}, "items": [{"bar_id": "901"}, {"bar_id": "902"}], "dry_run": true}}
{"tool": "manage_idea", "arguments": {"action": "create", "name": "Mobile app improvements"}}

The server uses a clean layered architecture:

┌──────────────────────────────────────────────────────────────┐
│                        cmd/productplan                        │
│                     (entry point, DI)                         │
└──────────────────────────────────────────────────────────────┘
                              │
        ┌─────────────────────┼─────────────────────┐
        ▼                     ▼                     ▼
┌───────────────┐    ┌───────────────┐    ┌───────────────┐
│  internal/cli │    │  internal/mcp │    │internal/tools │
│  (CLI cmds)   │    │  (MCP / SDK)  │    │  (handlers)   │
└───────────────┘    └───────────────┘    └───────────────┘
                              │                     │
                              └──────────┬──────────┘
                                         ▼
                              ┌───────────────────┐
                              │   internal/api    │
                              │  (HTTP client)    │
                              └───────────────────┘
                                         │
                                         ▼
                              ┌───────────────────┐
                              │  ProductPlan API  │
                              └───────────────────┘

Key interfaces:

// Tool handler interface (internal/mcp)
type Handler interface {
    Handle(ctx context.Context, args map[string]any) (json.RawMessage, error)
}

// Logging (internal/logging): a *slog.Logger with a JSON handler on stderr
logger := logging.New(slog.LevelInfo)

Logging format:

{"ts":"2026-09-24T10:30:00.123456789Z","level":"debug","msg":"API response","endpoint":"/roadmaps/5","status_code":200,"dur_ms":245}