GoKawiilAgentRun launches as open-source platform adding memory and recovery to AI agents
AgentRun is a newly released open-source project that layers self-improvement, durable execution and a management console on top of existing agent frameworks like Claude Code, Codex, OpenCode or pi. It lets agents retain notes, skills or workflows from past runs so they avoid repeating the same research, and it persists run state so interrupted sessions can resume rather than restart. The project is licensed under Apache-2.0 and can run locally via a single command, in a container, or as a server with runners.
GoKawiil's interpretation of the reporting above, not reported fact.
By having agents retain learned shortcuts rather than relearn tasks each time, the project's creators suggest users could cut token usage and costs on repetitive agent workloads. The addition of durable execution and a control plane with spend caps and approvals also points toward making autonomous agents more reliable and auditable for production use, though actual performance gains would depend on real-world testing beyond the demo data shown.
- AgentRun adds memory, durable execution and oversight tools on top of existing agent harnesses.
- It is open-source under Apache-2.0 and can run locally, in a container, or as a server.
- The stated goal is reducing repeated work and token costs across agent runs over time.
Source: github.com, 2026-11-12
Published there as: “Show HN: AgentRun – open-source, Self-improving Managed Agents”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.