Ensemble mcp User Documentation

Ensemble mcp is a Python MCP server providing vector memory, drift detection, model routing, skills discovery, session management, codebase indexing, context compression, and a local web dashboard for AI-assisted development pipelines.

All processing is 100% local — ONNX Runtime embeddings (~5ms), numpy cosine similarity, SQLite storage. Zero LLM or cloud API calls.

For Users

Get started with installing and configuring Ensemble mcp for your AI coding tool.

Guide	Description
Getting Started	5-minute quick start — install, register, verify
Installation	Detailed install: pip, source, Docker, system requirements
AI Tool Compatibility	Which AI tools work, what gets installed, which MCP tools each uses
CLI Reference	All commands: `serve`, `web`, `install`, `uninstall`, `add-agents`, `add-skills`
Configuration	Config files, layering, all settings with defaults
MCP Client Setup	Per-tool registration: OpenCode, Claude Code, Copilot, Cursor, Windsurf, Devin CLI
Web Dashboard	Dashboard usage, features, and JSON API endpoints
Troubleshooting	Common issues, error codes, and fixes

For Developers

Integrate Ensemble mcp into AI agent pipelines or contribute to the project.

Guide	Description
Tool Reference	All 19 MCP tools: parameters, types, response schemas, examples
Integration Guide	Pipeline patterns: pre/mid/post pipeline tool usage
Architecture Overview	System design, subpackages, data flow, extension points

Quick Links

Install: pip install ensemble-mcp
Register: ensemble-mcp install
Dashboard: ensemble-mcp web
GitHub: LynkByte/ensemble
Issues: GitHub Issues
Changelog: CHANGELOG.md

19 MCP Tools at a Glance

Category	Tools	Purpose
Patterns	`patterns_search`, `patterns_store`, `patterns_prune`	Semantic memory of past solutions
Drift	`drift_check`	Detect scope drift during implementation
Routing	`model_recommend`	Choose model tier per agent and task
Skills	`skills_discover`, `skills_suggest`, `skills_generate`	Find, suggest, and create reusable skills
Session	`session_save`, `session_load`, `session_search`	Pipeline checkpoints with resume support
Indexer	`project_index`, `project_query`, `project_dependencies`, `project_snapshot`	Codebase intelligence
Compress	`context_compress`, `context_prepare`	Token-efficient context optimization
Utility	`health`, `reset`	Server status and data management