Infinite Context

Infinite context for your AI model.
Unlimited memory for coding agents, without putting it in the context window.

An agent using Infinite Context never loses a fact, and loads almost nothing on any given turn. Every prompt gets **five pointers**, a slug and a one-line description each, and the agent opens only the one or two memories the question actually needs. The memory can grow to thousands of facts. The prompt stays the same size.

It runs entirely on your machine. No account, no cloud, no data leaving the building.

Quick start

Requires Node 18 or newer. Nothing else.

git clone https://github.com/Taha-ElBouzidi/infinite-context
cd infinite-context
node tools/init.mjs            # seeds five rules, installs the embedding runtime, builds the index
node tools/install-hook.mjs    # wires Claude Code to it. Restart your session after.

That is the whole install. Recall works on the next prompt, by keyword **and by meaning**: a question with no word in common with a memory still finds it. The first `init` downloads the embedding model, about 280MB, once. The hook starts the local emb