Ona is a startup working on the coordination layer for agentic software development — the “missing primitive” that makes it possible to reliably orchestrate, constrain, and advance many AI agents through the software development lifecycle (SDLC).
The company’s research centers on the observation that while runtime, orchestration, and triggers for multi-agent systems are mostly tractable, coordination between agents and across workflow stages remains the hardest unsolved problem. Ona approaches this through what founder Lou Bichard calls “harness engineering”: encoding feedback loops into the repository itself through specs, agents.md, tests, and skills, so agents can self-correct without constant human intervention.
Ona’s practical work includes developing secure, VM-based execution environments for agent swarms (favoring VMs over containers for stronger isolation), designing explicit state machines and durable workflow-style execution patterns, and building CLI gateway patterns that tell an agent whether it has completed a SDLC stage and may proceed. The company is building in a still-early ecosystem where most coordination solutions are internal, fragmented, and not shared — making their work on standardizing the coordination layer particularly valuable for teams scaling agentic software factories.