PolyPlan MCP

A multi-model AI planning MCP server for CLI coding tools

What is PolyPlan?

PolyPlan is a project-local MCP (Model Context Protocol) server that enables developers to run structured, multi-model AI planning sessions across multiple CLI coding tools simultaneously.

When developers use multiple AI models (Claude Code, Copilot, OpenCode, Codex, Cursor, AntiGravity) for planning, PolyPlan provides a structured way to collect, cross-review, and synthesize plans from all models into one final implementable master plan.

How It Works

PolyPlan orchestrates a 3-round planning workflow:

• Round 1: Each model creates an independent plan saved to .plans/ without seeing what any other model thinks.
• Round 2: Each model reads all other models' Round 1 plans (except its own) and creates a revised master plan informed by peer review.
• Final: One chosen model (your strongest, e.g., Opus 4.6 or GPT-4o) reads ALL Round 1 + Round 2 plans and produces the final, synthesized, implementable plan.

Installation

npm install -g polyplan-mcp

Note: Global install does best-effort user-level MCP registration. Run polyplan-mcp init inside each project to create project-local MCP config and slash-command wrappers.

Quick Start

1. Initialize in your project

   polyplan-mcp init

2. Restart your AI coding CLI/editor
3. In any connected CLI tool, start Round 1

   /polyplan round1 [your problem]

   Or use the direct command:

   /round_1 [your problem]

4. Check status anytime

   /polyplan status

   Or:

   /show_status

All Commands

These are the MCP tool names and the generated slash wrapper names. PolyPlan keeps MCP tools as the execution layer and creates client-specific prompt/command files during polyplan-mcp init where supported.

| Slash Command | Description | |---|---| | /round_1 | Start Round 1 — create an independent plan (no other models seen). | | /round_1_context | Get the Round 1 prompt — call before generating your plan. | | /round_2 | Start Round 2 — peer-review all Round 1 plans and write a revised master plan. Requires ≥2 Round 1 plans. | | /round_2_context | Get the Round 2 prompt with all other models' plans injected. | | /final_plan | Start Final round — synthesize ALL plans into one implementable plan. Requires ≥1 Round 2 plan. | | /final_plan_context | Get the Final round prompt with all Round 1 + Round 2 plans injected. | | /show_status | Show full session state — which models completed each round. | | /clear_plans | Delete plan files. Specify all, round1, round2, or final. Requires confirm=true. | | /show_conflicts | Show all points where models disagreed in Round 1. | | /show_questions | Show all open questions raised by any model, and which were answered by others. | | /show_diff | Show what changed for a specific model between Round 1 and Round 2. | | /show_agree | Show what ALL models agreed on independently in Round 1. | | /show_summary | One-paragraph summary of each model's plan. | | /export_plans | Bundle entire .plans/ session into one readable markdown file for sharing. | | /show_history | Full audit log — model, CLI tool, round, time, action. |

PolyPlan also generates /polyplan as a routing command where client command files are supported:

/polyplan status
/polyplan round1 [problem]
/polyplan round2
/polyplan final

File Naming Convention

All plans are stored locally in the .plans/ directory using the following convention:

.plans/
  round1-copilot-sonnet4.6.md
  round1-claudecode-kimik2.md
  round2-opencode-gemini2.5.md
  final-antigravity-opus4.6.md

Supported CLI Tools

PolyPlan connects to any tool supporting the Model Context Protocol (MCP):

• Claude Code (.mcp.json, .claude/commands/)
• Cursor (.cursor/mcp.json, .cursor/commands/)
• OpenCode (opencode.json, .opencode/commands/)
• VS Code / GitHub Copilot Agent mode (.vscode/mcp.json, .github/prompts/)
• GitHub Copilot CLI (.mcp.json)
• Codex CLI (global ~/.codex/config.toml best-effort postinstall registration)
• Windsurf / AntiGravity / Gemini CLI workflows (MCP tools and prompts where configured)

See COMPATIBILITY.md for the full client support matrix and limitations.

CLI Tool Specific Usage

Always pass your model name

PolyPlan saves plans using the model name in the filename (e.g., round1-claudecode-sonnet4.6.md). Auto-detection is not always possible, so always pass modelName explicitly:

Call the round_1 tool with modelName='sonnet4.6', problemDescription='...', and plan='...'

OpenCode / Gemini

OpenCode with Gemini models requires explicit tool invocation syntax. Use "Call the tool" instead of "Use polyplan":

| ❌ May not invoke the tool | ✅ Always invokes the tool | |---|---| | Use polyplan round1 for this problem: ... | Call the round_1 tool with modelName='gemini2.5' and problem='...' and plan='...' | | Use polyplan to show status | Call the show_status tool | | Use polyplan round2 | Call the round_2_context tool, then call round_2 |

The word "Call" forces direct MCP tool invocation in OpenCode/Gemini instead of a natural-language response.

Claude Code

Claude Code auto-detects as claudecode but does not expose the active model name via MCP. Always pass modelName explicitly:

/round_1 modelName=sonnet4.6 problemDescription="..." plan="..."

GitHub Copilot (VS Code)

Works with slash commands directly. Still recommended to pass modelName:

/round_1 with modelName='sonnet4.6' for this problem: ...

Why Multi-Model Planning?

Different models have different strengths, blind spots, and reasoning styles. Some excel at architectural structure, while others are better at catching security edge cases. By using 5+ models independently and then cross-reviewing, you catch more issues, resolve conflicts early, and produce a significantly more robust plan than any single model could produce alone.

Contributing

We welcome contributions! Please see our CONTRIBUTING.md for details on how to get started, run tests, and submit Pull Requests.

License

MIT © 2026 PolyPlan Contributors