Synaptic documentation
Quickstart
This walks through building a graph for a repository and using it. It assumes the
synaptic binary is on your PATH (see Installation).
1. Build the graph
cd your-project
synaptic extract .
This scans the directory, extracts symbols and relationships, clusters and analyzes the
graph, and writes everything to synaptic-out/. It honors .gitignore and
.synapticignore, skips dependency and build directories, and skips sensitive files such
as .env and private keys. A code-only run makes no network calls. See
Extraction for the full discovery and caching behavior.
Key outputs (full list in Output Formats):
synaptic-out/graph.json- the canonical graphsynaptic-out/GRAPH_REPORT.md- structural insight reportsynaptic-out/graph.html,graph-3d.html,graph.svg- visualizations
2. Read the report
Open synaptic-out/GRAPH_REPORT.md for god nodes (most-connected symbols), surprising
connections, import cycles, and suggested questions. See
Analysis and Reports.
3. Query the graph
# Relevant subgraph for a topic:
synaptic query "authentication flow"
# Depth-first expansion (favors deep call chains):
synaptic query "request lifecycle" --dfs --max-nodes 40
# Shortest path between two symbols:
synaptic path login validate_token
# A node and its neighbours:
synaptic explain parse_config
# Reverse impact: what depends on a symbol?
synaptic affected parse_config --depth 3
See Querying for how seeds are resolved and how traversal works.
4. Explore visually
Open the generated files in a browser:
graph.html- 2D explorer with search, community color, and relation filtersgraph-3d.html- 3D force graphtree.html- file tree;callflow.html- call-flow diagram
See Visualizations.
5. Keep it current
synaptic update path/to/changed_file.rs # incremental rebuild
synaptic watch # rebuild automatically on save
synaptic hook install # rebuild on commit/checkout via git hooks
See Incremental Updates.
6. Use it from an AI assistant
synaptic serve # MCP server over stdio
synaptic install # wire Synaptic into a host assistant
install wires Synaptic into a host assistant so it queries the graph before grepping or
reading files: a Claude skill file + hooks, an always-on instructions block for other
assistants, and a native MCP server + hook for Codex. See MCP Server and
Assistant Integration.
Next steps
- Multiple repos or a monorepo: Workspaces and Federation
- Enrich docs/papers with an LLM: Semantic Analysis
- Pull in external sources: Ingestion
- Every flag of every command: Commands