What is codebase memory for AI coding agents?

Structure: which files exist, which functions call which, what imports what, which tests cover which code. This is what lets an agent find the right place to change without searching.

History: why code is the way it is. The pull request that added a strange retry, the review comment that asked for it, the revert that followed. This is what stops an agent from “cleaning up” code that keeps production running.

Without memory, every task starts with exploration, which costs tokens and time, and the agent still can't see intent that isn't written in the code. With memory, it starts from a brief: the files to touch, the decisions not to break, and the blast radius of a change.

Synapse builds a code graph from your GitHub repository (TypeScript, JavaScript and Python) and links the why from pull requests, reviews and commits to the exact lines. AI tools reach it over MCP; a PR bot adds the same context to every pull request. It is read-only and never trains on your code.

In our benchmark (Claude Code with Sonnet, 7 real tasks on a repository Synapse had never seen, 2 runs each), Synapse cut tokens by 58% and cost by 48%, and every edit still compiled. Planning tasks saved the most. Results vary by repository and task.

Synapse is free to start: connect a GitHub repository and add it to Claude Code, Cursor or Claude in about two minutes. Create an account or read the docs.