A desktop app that runs several Claude Code and Codex agents at once, each in its own git worktree, with a gate that proves the work before it counts as done.
Lab. Everything we ship
Three open products and a free directory, built for our own work and released for anyone else’s. We sell the expertise that built them as audits and build sprints; the products come first.
A free subdomain registry: claim yourname.runs-on.dev with a GitHub account, and the claim itself is one git commit to a public repository.
An open-core suite that checks whether AI answer engines can find, parse and cite a site: browser tools, a Chrome extension and three hosted MCP servers.
A free directory of companies building AI-agent and MCP tooling: one page per company, listing every MCP server it publishes to the Official MCP Registry.
Two lines of research
Urban Drone Autonomy
Sim-first mission autonomy for a multirotor: a behaviour state machine with a safety layer holding veto power, 3D navigation through real cities imported from OpenStreetMap, and camera perception that infers powerline catenaries from pole detections. 275 tests, and a first PX4 SITL flight (takeoff, hold, return, land) that proves the backend seam.
Quantum-hybrid
A public research repo pairing quantum and classical methods, with a claim-tag taxonomy and a 72-test emulator so a result is either verified or labelled as unverified — never left ambiguous.
The parts, published separately
8 packages on npm and one Chrome extension, all Apache-2.0. The audit we sell is built on the same scoring engine you can install.
Everything above is open source on one GitHub org, except the AEO Toolkit's hosted console and MCP servers, which are proprietary.