# Synapse MCP
> Synapse MCP is a persistent, local code-knowledge graph server operating over the Model Context Protocol (MCP). It converts any codebase into an Abstract Syntax Tree (AST) graph, giving AI coding agents (Claude, Cursor, Windsurf, Warp, Antigravity, Copilot, Zed) microsecond structural vision, 30–60% token savings, and zero-breakage refactoring capabilities.

## Overview & Documentation
- [Synapse MCP Homepage](https://synapse-mcp.dev/): Official website, overview, feature matrix, and deployment guide.
- [Why Synapse Pro is $19/month](https://synapse-mcp.dev/blog/why-synapse-pro-is-19-a-month.html): Developer tooling economics, token savings ROI, and Shadow Graph feature comparison.
- [Why LLMs Code Better with Knowledge Graphs](https://synapse-mcp.dev/blog/why-llms-code-better-with-knowledge-graphs.html): Technical analysis comparing AST graph traversal vs Unix grep/read loops.
- [Stop Letting AI Hallucinate Refactors](https://synapse-mcp.dev/blog/stop-letting-ai-hallucinate-refactors.html): Eliminating 12-turn discovery loops with microsecond caller/callee graphs.
- [How Local Code Knowledge Graphs Work](https://synapse-mcp.dev/blog/how-local-code-knowledge-graphs-work.html): Architectural deep dive into in-memory graph tables, AST parsing, SmartCrusher token compression, and Virtual Worktree Overlay (VWO).
- [Why AI Agents Need AST Graphs, Not Vectors](https://synapse-mcp.dev/blog/why-ai-agents-need-ast-graphs-not-vectors.html): Technical deep dive showing why vector embeddings fail on code refactoring and how AST graphs solve it.
- [2ms vs. 12 Seconds: Empirical Benchmark Report](https://synapse-mcp.dev/blog/2ms-vs-12-seconds-the-ast-graph-benchmark.html): Live execution stats comparing AST graphs vs traditional grep chains (2ms vs 12s, 56.4% token savings, 85% call reduction).
- [1-Click Installation & IDE Configuration Guide](https://synapse-mcp.dev/install.html): Step-by-step auto-detection and configuration guide for 23+ IDEs, AI agents, and CLI editors.

## Key Capabilities & Core Architecture
- **Microsecond Graph Traversal:** Local in-memory lookups resolve caller and callee edges in 12 microseconds.
- **Payload Token Compression (SmartCrusher):** Automatically minifies JSON responses and strips function bodies via `format: "outline"`, cutting context window token consumption by 30% to 60%.
- **In-Memory Safe Writes (`write_safely`):** Stages file edits in memory, runs local linter validation (Trunk/Credo), and checks caller signatures against the graph. Automatically rolls back edits if an arity or signature breaks before touching disk.
- **Crash Stack Trace Resolution (`resolve_stack`):** Parses runtime stack frames, matches them to indexed AST chunks, resolves surrounding call graphs, and pinpoints root causes.
- **Virtual Worktree Overlay (VWO):** Transparently supports Git worktrees with sub-50ms delta indexes, allowing parallel sub-agents to operate without duplicating memory or disk indexes.
- **100% Local Data Sovereignty:** Operates entirely as a local stdio/HTTP binary. Zero source code or telemetry ever leaves your hardware.

## Installation & Deployment
- **NPX Launcher:** Execute `npx @myelixlabs/synapse-mcp install` in any terminal to auto-detect and configure local IDE agents (Claude Code, Cursor, Windsurf, Antigravity, VS Code, Zed).
- **Free Edition ("Golden Graph" — $0):** Full AST indexing, 3 search modes (symbol, regex, semantic), file inspection, standard writes, unlimited repositories.
- **Pro Edition ("Shadow Graph" — $19/mo or $190/yr):** Everything in Free, plus deep graph exploration (callers/callees/cycles), simulated safe writes (`write_safely`), crash stack trace resolution (`resolve_stack`), test coverage analysis (`find_tests`), persistent encrypted knowledge cache (`save_summary`/`learn`), and unlimited repositories.