Graph-backed code intelligence for your AI assistant.
Constellation turns your codebase into a queryable knowledge graph, giving AI assistants the structural understanding they need to reason about real software — not just the plain text.
Why Constellation?
Text search tells you where a string appears, everywhere that string appears. Constellation tells you the exact location of the symbol in question, what it means, what calls it, and what breaks if you change it. Before your assistant edits a function, it can ask:
- Where is this defined, and where is it used across the codebase?
- What's the blast radius of this change?
- Which modules have circular dependencies or dead code?
- How does data flow through the call graph?
Answers come from a semantic graph, not a grep loop.
One Tool, Countless Capabilities
A single code_intel tool exposes a rich JavaScript API as a "Code Mode" tool. Your assistant composes queries like code:
const { symbols } = await api.searchSymbols({ query: "AuthService" });
const [impact, deps] = await Promise.all([
api.impactAnalysis({ symbolId: symbols[0].id }),
api.getDependents({ filePath: symbols[0].filePath })
]);
Built for Real Codebases
- Multi-language: TypeScript, JavaScript, Python, and more coming soon.
- Multi-project: Auto-resolves per-project config by git root with LRU caching.
- Fast: Parallel API calls, rich source snippets returned inline, no follow-up file reads needed.
- Safe: Hardened sandbox mode runs queries in an isolated child process with hard memory limits.
Privacy by Design
Your source code never leaves your machine. Constellation parses to a compressed AST locally; only structural metadata reaches the service. Access keys scope each project to its own namespace.
Server Config
{
"mcpServers": {
"constellation": {
"command": "npx",
"args": [
"-y",
"@constellationdev/mcp@latest"
],
"env": {
"CONSTELLATION_ACCESS_KEY": "${CONSTELLATION_ACCESS_KEY}"
}
}
}
}