Why we're doing this
Over the last year, one thing has become clear to us: agents aren't a smarter search box — they're a new kind of colleague. But the substrate that would let them actually take the work barely exists. What's out there is either a polished chat box or a few lines of prompt against a model API.
Doing the work ourselves made it obvious: the next generation of organizations won't be decided by who has the strongest model. It'll be decided by who first builds the substrate underneath the human-plus-agent team — long-lived context, on-record work items, runtimes you actually own, compounding skills. That's infrastructure, not a feature.
GracePeak Labs exists for that. We're not trying to ship another AI product. We're trying to make sure anyone who wants to bring agents into their organization as first-class peers has a reliable base layer to build on.
A few things we won't compromise on
People are the protagonists
Agents are colleagues, not stand-ins. Judgment, taste, and decisions that matter stay with people.
Real books, kept open
Every step an agent takes has to be on the record, replayable, and revocable. Invisible work is not trustworthy work.
Your substrate, your keys
Agents run on your own machine with your own sessions. We don't want our cloud holding the key to your work.
Ship the work
The point isn't how fluently it chats. The point is whether it carries the work all the way through.
One open base, for anyone
All the infrastructure agents need to actually take the work — built as an open, general-purpose base so anyone can pick it up.
The product you're on now. A general base for agents that actually take the work: long-lived context, work items, your own runtime, compounding skills, browser-as-tool — for anyone bringing agents into a team.
Who's doing it
A small, hands-on team across engineering, product, and AI architecture. We're the tool's authors and its earliest heaviest users.
- 01
Founder
Holds the direction, the judgment, the trade-offs that matter. Also a primary author of the early product.
- 02
Engineering
Builds the substrate. Runtimes, long-lived context, work items, CLI — the layer that lets agents actually ship the work.
- 03
Design
Tools people want to keep using. Folds the complexity into surfaces with a clear sense of touch.
- 04
AI Architect
Large language models and agentic workflows — making the human-plus-agent collaboration actually run.
Want to talk?
Want to try it, partner with us, join us, or just curious — we'd like to hear from you.