axon

Axon builds a structural knowledge graph of your codebase and exposes it through MCP tooling so AI agents can reason about code relationships, dependencies, and execution flows without parsing flat text. Once indexed with axon analyze <path>, Axon provides powerful queries and analyses via commands such as axon analyze, axon ui, and axon context/impact-like tools. Use the MCP server to serve a local graph explorer and programmatic endpoints that return contextual results (e.g., which symbols are impacted by a change, or how a symbol participates in an execution flow). The included features let you search by name or concept, inspect callers and callees, detect dead code, and explore communities and execution flows, all with structured, graph-backed responses rather than raw file diffs.

To use Axon within the MCP workflow, first index your repository with axon analyze <path>. Then run the MCP server (for example, via the provided axon command) to start serving the graph and web UI routes. You can pair it with the Axon UI for an interactive visual exploration surface or query endpoints that return grouped results by execution flow, confidence scores, and community membership. The tools are designed to keep context dense in responses, so agents spend tokens reasoning about architecture rather than navigating flat text.

Common use patterns include: (1) performing an impact analysis to surface affected symbols when editing a given symbol, (2) performing a flow-aware search to locate related code regions organized by execution paths, and (3) inspecting dead code and network of dependencies to prioritize refactoring with architectural awareness.

Prerequisites

• Node.js (recommended: LTS version) and npm installed on your system
• Access to a terminal/command prompt

Installation steps

1. Install Axon (CLI/server) globally via npm:

npm install -g axon

2. (Optional) Initialize or upgrade your project so Axon can index a repository:

cd /path/to/your/project

3. Index your codebase (first-time run):

axon analyze .

This builds the knowledge graph: parses files, traces calls, detects communities, and stores embeddings.

4. Start the MCP server to serve the graph and UI:

axon serve --watch

This serves the MCP endpoints and a web UI at the configured port (default localhost:8420) with live-reload during changes.

Notes

• You can customize the port and dev options for the UI via additional flags (see Axon docs).
• Ensure your project has read access to the files you want to index.

Tips and considerations:

• After indexing, Axon updates the graph in watch mode so changes are reflected in near real-time.
• Use axon ui to explore the knowledge graph visually; the UI exposes Explorer, Analysis, and a Cypher Console for advanced queries.
• For best results, index a representative subset of your codebase first to validate the tooling before scaling to larger repositories.
• If you encounter performance issues on very large repos, consider indexing in incremental passes or filtering to relevant modules first.
• Environment variables can be used to configure runtime paths or network ports if you’re deploying in containers or CI environments; consider setting AXON_PORT, AXON_WATCH, or related vars per your deployment setup.