Open Source Agent Tools: November 13, 2026
Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled
1. AgentRun Is Open-Sourced Under Apache 2.0 as a Self-Hostable Managed-Agents Layer That Adds Learning and Durable Runs to Existing Harnesses
AgentRun (Parcha-ai/agentrun, Apache-2.0) was announced on Hacker News on November 12 by GREP.AI (YC F26), which says it is open-sourcing the platform it built for its own agent harness and turning it into a managed agents product. It does not replace the agent loop: it wraps Claude Code, Codex, OpenCode, pi or any harness that speaks ACP and adds a control plane with approvals, spend caps and per-run cost tracking. After each run the agent keeps a note, a skill or a workflow that can do the job without the model loop, and every run is persisted in one store so a run resumes after the server is killed and restarted. Agents can be given a sandbox, a Linux desktop, an Android phone or a browser as their environment. It starts locally with npx @parcha/agentrun (Node 24.22.3 or newer, macOS or Linux), deploys as one container or as a server with runners, and is also offered as a hosted console; the latest tagged build is v0.1.0-beta.35. The README’s cost claims, including a code-review agent at 18 cents a review that caught 1.6 times the real bugs of a commercial reviewer on 30 unseen pull requests, are the company’s own case studies and have not been independently reproduced. Repository Source