devplan-mcp-server

DevPlan MCP Server is an MCP server designed for Claude Code, aimed at turning project ideas into step-by-step executable development plans. It is based on the ClaudeCode-DevPlanBuilder methodology and solves the problems of AI programming assistants often losing context, skipping steps, or producing inconsistent code across sessions.

Main Capabilities

• Generate development plans: Through devplan_generate_plan create the foundational phases and feature list, and generate files such as PROJECT_BRIEF.md, DEVELOPMENT_PLAN.md, CLAUDE.md, etc.
• Validate plan executability:devplan_validate_plan check whether the plan is truly executable, including complete code blocks, import statements, no "add to existing" instructions, no cross-subtask references, validation commands, and expected output.
• Multi-model support: Through the target parameter, generate adapted output for different AI programming tools, supporting Claude, Cursor, Aider, Cline, Windsurf, and generic formats.
• Real-time progress: The generated executor and validator agents integrate with Claude Code's Task tool to display subtask progress in real time.
• Lessons learned system: Capture issues from validation and incorporate them into future plans.
• Issue fixing: Turn GitHub Issues into structured fix tasks.
• Dashboard and analytics: Provide a public dashboard to view aggregate usage statistics, with privacy protection, no IP storage, anonymous sessions, and daily statistics expiring after 90 days.

Usage

Install as a Skill (recommended, no MCP server required)

1. Add the repository as a plugin marketplace:

   /plugin marketplace add mmorris35/devplan-mcp-server
   

2. Install the plugin at the user scope:

   /plugin install devplan@devplan-mcp-server --scope user
   

3. Reload the plugin:

   /plugin reload
   

4. Invoke the skill directly:

   /devplan
   

   Or use a subcommand, or directly describe the requirement, and Claude will invoke it automatically.

Install as an MCP server

Under ~/.claude.json in mcpServers , add the configuration under the key.

Update an existing installation

Remove from both scopes and then add again.

Workflow

DevPlan uses a "scaffold -> enhance -> validate" workflow:

1. Interview: DevPlan asks questions to understand the project.
2. Brief: Create a structured PROJECT_BRIEF.md.
3. Generate scaffold:devplan_generate_plan create the foundational phases and feature list.
4. Code enhancement: The AI agent structures the feature phases and fills in complete copyable code.
5. Validation:devplan_validate_plan check plan executability.
6. Execution: The executor agent implements each subtask, including inline git commands.
7. Validation: The validator agent tries to break the implementation.
8. Learning: Issues become lessons learned for future projects.

Applicable Scenarios

• Developers who need to turn project ideas into detailed, executable development plans.
• Those using Claude Code or other AI programming tools for project development and wanting to maintain context consistency and step completeness.
• Those needing multi-model support to generate adapted plans for different tools.
• Those wanting to learn from validation and improve future plans.
• Those needing to turn GitHub Issues into fix tasks.

Limitations

• As an MCP server, it depends on the network and may experience SSE timeouts; the Skill form has no network dependency and is recommended.
• In multi-model support, different target adapters have differing capabilities; see docs/ADAPTERS.md for details.
• Dashboard analytics only provides aggregate statistics and does not involve user identification.
• Specific fees, pricing, platform compatibility, etc. are not provided in the materials.