AgentRun
Open-source platform for self-improving AI agents
Visit website →About
AgentRun is an open-source platform that adds self-improvement, durable execution and a control plane on top of your favorite agent harness. It works with Claude Code, Codex, Pi and OpenCode.
After each run, the agent keeps what will help next time: a note, a skill, or a workflow. The next run starts from there instead of rediscovering the same API or rebuilding the same plan. Every run lives in one store. Kill the server mid-run, start it again, and the run carries on.
You stay in control with approvals, spend caps and one console for every agent. AgentRun is Apache-2.0 licensed and runs on your own infrastructure with your own model keys.
AI-assisted draft, human-reviewed before publishing — see how we choose & review tools.
Why we picked it
Most agent platforms treat every run as a blank slate. AgentRun remembers what worked, turns repeated steps into reusable skills, and cuts token waste on tasks your agent has done before. Open-source and harness-agnostic — it wraps Claude Code, Codex, Pi, and OpenCode in one control plane.
Related tools
- 1 APC Security for AI agents: contracts bound to each action Custom
- 2 HUMAIN ONE HUMAIN's enterprise agent platform, announced with Turing, AWS, Microsoft Custom
- 3 Agentic Identity Broker OAuth2 delegation and consent management for AI agents Free
- 4 Nico Voice-to-action agent for Mac, runs locally Freemium
- 5 phpClaw Open-source AI agent framework for PHP with tools, memory and guards Free
- 6 BotGauge Red-team your AI agents and find failures before your users do Freemium