01
Local by default
The raw archive lives on machines you own. Anything sent to a model provider is a deliberate, bounded step.
LOCAL-FIRST MEMORY FOR AI TOOLS
Memory your AI
tools can cite.
Tracewell keeps the conversations you have with AI assistants and coding agents on your own machines, and lets any of them find what was said and decided, with the original source attached.
In daily private use · Public release in preparation
01 / PRODUCTOne place to ask
One place to ask
“what did we decide?”
MEMORY & CONTEXT LAYEREvery tool forgets.
Every tool forgets.
Your archive doesn’t.
Work is spread across chat assistants, coding agents and terminals, and each one starts from zero. Tracewell archives those sessions locally and puts a single search command in front of them, so the next assistant can pick up a decision made last month in a different tool, and show you the conversation it came from.
Cross-tool session archiveOne search entry pointSource-linked recallRetrieval benchmark
Chat & agent sessions
Tracewell Archive · Search · Cite
Any AI tool you use next
02 / PRINCIPLES
Built around trust.
02
Sources over summaries
A distilled memory is a claim until it points back to the conversation it came from. Recall returns the evidence, not just the conclusion.
03
Measured, not assumed
Retrieval is tested against a private benchmark of 206 real questions over 87 past sessions, so changes are judged by whether the right conversation is found.
03 / STATUSWorking today.
Working today.
Opening up next.
Today Tracewell runs as a command-line front door over a local session archive, with hourly backups and a retrieval benchmark. Next: distilled memories with evidence links, a stable interface other AI clients can call, and an installable public release.
Tracewell builds on open-source foundations, including AgentsView and hstry (both MIT-licensed), for session indexing and conversation capture. Questions or early access: hello@tracewell.dev