Skip to content
We’re going live Sunday, April 26 at 9:00 PM CT.
About

A tool for the engineer who has hundreds of Claude Code sessions and cannot find any of them.

You started with Claude Code because the leverage was real. Three months in, you have more sessions than you can name. The agent is brilliant inside a single conversation and a stranger across them. You re-explain your project every morning. You search by opening folders. You feel the productivity tax even when you cannot quite measure it.

That tax has a name now. It is called context debt. Claude Recall is the tool that pays it off.

What this is for

Claude Recall indexes every Claude Code session you have ever run into a local SQLite database, makes it searchable, and lets you pipe any past session back into a fresh conversation as starting context. The agent stops forgetting. You stop re-explaining. The tool lives entirely on your machine; no cloud, no telemetry, no signup.

If you have crossed the hundred-session line and felt the wall, this is the tool. If you have not, you will. Claude Code usage is growing six-x year-over-year, and most users hit the wall at month three.

Why local-first

The conversations you have with Claude include your code, your customer data, your half-formed business decisions, your debug trail. Sending those to a server is a deal you should not have to make. Every line of Claude Recall is engineered around the opposite assumption: nothing leaves your machine, ever. The daemon binds to 127.0.0.1 only. The dead-man clause says a future version of Claude Recall can never delete your past, including a future version shipped by us.

The market is full of memory tools that hold your data hostage in exchange for search. That is a trade-off you do not have to take.

Who is behind it

Claude Recall is built by one developer, in public, for a community of people running ten or twenty terminals at a time. Solo means the feedback loop is short: a Claude Code user posts a real frustration on Discord; a fix ships in the next release; the changelog links back to where it came from. There is no roadmap committee. Every feature traces to a session that hit a wall.

Solo also means we promise less than the larger players. There is no SOC 2 audit. There is no enterprise SSO at v0.6. There is no team-shared memory yet. What there is: a tool that was built because the maker had the same problem you have, and could not find an answer he trusted.

What we will not do

Where to start