
MapX
Local code graph memory for LLMs
2 followers
Local code graph memory for LLMs
2 followers
Local code graph memory for LLMs: scan a codebase once, then query symbols, dependencies, impact and architecture rules without re-reading files.

Free
Launch Team

CloudflareDeploy serverless functions globally in seconds
Promoted
Scanning once and querying instead of re-reading files is exactly the right trade-off — that's where most of the wasted tokens go in a long session. The impact queries are the part I'm most curious about. How does the graph stay fresh as the code changes — is it an incremental rescan, or do you re-scan on demand? And which languages does it support at launch?
@jcbuilds Impact and blast-radius is determined by tracing dependencies and dependents class/file references. MapX prints warnings in both CLI and MCP tools' response if graph is stale to inform agents to re-scan the codebase, which only parses new file changes keeping the graph fresh or after switching branches. MapX is git-aware and multi-repositories / mono-repo friendly. It supports 23 languages and 22 frameworks, also comes with 39 MCP tools.
Check MapX website for more info: https://mohamedgamil.github.io/MapX/