About Ouroboros
Ouroboros is an open source general purpose AI agent that keeps a persistent identity, memory and history across tasks and restarts, and can modify its own implementation. It works on external projects in separate workspaces, coordinates a swarm of specialist agents, and can change its own code, prompts, tools, architecture and dependencies. Protected constitutional and safety files carry stricter rules, and git history, task records, tests and review evidence keep changes inspectable.
It is aimed at developers and researchers who want an autonomous agent they can run locally. One root agent stays responsible for each task and its final result, while specialists research, build, test and challenge the work in parallel lanes. It runs as a native desktop app or through a headless command line. The runtime, repository, memory, logs and task history stay on the user's machine, and model inference uses remote APIs the user configures or a local GGUF model.
The vendor publishes self reported results on Terminal Bench 2.1, OSWorld Verified and CL Bench, along with the evidence, and notes a statistical tie with Codex CLI on a SWE bench Pro matched pair. These are the vendor's own figures.
No pricing is published. Because model calls go to providers the user configures, usage costs depend on the chosen provider.