# Synaptic Code Intelligence [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/ColinVaughn/Synaptic  
**GitHub Stars:** 31  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/synaptic-code-intelligence

## Description
Code intelligence for repository graphs, change impact, test selection, and engineering memory.

## Claude Desktop Quick Installation
Heuristic fallback — verify the package name and runner against the repository README before running it. Uses `npx` (confidence: low):

```json
"mcpServers": {
  "synaptic-code-intelligence": {
    "command": "npx",
    "args": ["-y","synaptic-code-intelligence"]
  }
}
```

## Documentation

## What Synaptic Code Intelligence MCP server does

Synaptic Code Intelligence MCP server builds a navigable representation of a repository and makes relevant portions available to coding assistants. The graph can include symbols, calls, imports, inheritance, resources, SQL, and relationships across language boundaries. It supports questions about how a subsystem works, what depends on a symbol, and which parts of a system may be affected by a proposed change.

The project also covers repository memory and maintenance workflows. Its documented capabilities include source-linked history from commits, decisions, incidents, reviews, CI, and earlier attempts; API contract tracking; vulnerability evidence; structural search; multi-repository federation; graph-aware pull-request review; and exports to GraphML, Cypher, Graphviz, Obsidian, and Markdown.

## How it works

Extraction is local and deterministic. Running `synaptic extract .` creates a `synaptic-out/graph.json` file, while `synaptic update` and `synaptic watch` keep the graph current. Extraction follows `.gitignore` and `.synapticignore`, skips common secret files, and remains offline unless a network-backed feature is explicitly enabled.

The MCP service starts with `synaptic serve`. Assistants can then retrieve focused graph slices instead of repeatedly reading full source files. The command-line interface also provides direct operations such as `query` for subgraph retrieval, `affected` for reverse-impact analysis, `predict` for risk and test selection, `speculate` for isolated-worktree verification, and `memory search` for repository history.

## Setup and configuration

Synaptic Code Intelligence MCP server is distributed as a single Rust binary. The README provides installation scripts for macOS, Linux, and Windows, and supports building from source with Rust 1.97.1. After installation, run extraction from the target repository before querying it or connecting an assistant.

Assistant setup includes `synaptic install claude` for Claude Code, `synaptic install codex` for Codex CLI, and `synaptic install codex --global` for the Codex desktop app. Gemini, Cursor, Copilot, OpenCode, Kilo, and generic `AGENTS.md` clients are also listed as supported. The excerpt does not provide client-specific configuration details for those integrations.

## Tools and capabilities

The Synaptic Code Intelligence MCP server is suited to workflows that need:

- Repository and architecture understanding through symbols and relationships
- Reverse-impact analysis for a symbol or proposed change
- Change forecasting and affected-test selection
- Optional verification in a throwaway worktree
- Search across source-linked engineering history
- Architecture charts generated with `synaptic chart`
- Incremental updates and federated multi-repository workspaces

## Limitations and notes

The provided benchmark reports that full-graph answers used fewer tokens than reading referenced source files, but it explicitly describes this as a context-compression measurement rather than a guarantee about total task cost. The repository is licensed under AGPL-3.0-or-later. The separately maintained Synaptic Cloud service is proprietary and is not covered by this repository license.

_Full upstream README: https://allmcps.com/mcp/synaptic-code-intelligence/readme_