Queue a local Claude Code test run from a sandboxed agent workflow
At a glance
A developer built a file-based bridge that lets a sandboxed agent ask Claude Code on a local machine to run and repair a test suite. This workflow is listed under local-execution, asynchronous-workflows, agent-tooling.
Agents / tools
Microsoft Agent Framework, Claude Code, cowork-to-code-bridge
The agent writes a task into a shared directory. A local daemon executes it and writes back results. The example queues run_claude.sh with a test-and-fix request, then polls by task ID so a long build can outlast a short sandbox session.
How it went
The author reports a working implementation with more than 330 tests and support for macOS, Linux, and WSL2.
Worth knowing
The bridge exposes per-task budgets, permission scopes, an approval option, and idempotency keys. The author is still seeking feedback on the best Agent Framework integration point.
Save ideas in this browser. Upvotes count across the board.
CircleCI's Chunk sidecars run scoped microbuilds in Firecracker/E2B microVMs that mirror CI, hooking agent stop/eval events so Claude Code, Codex, or Cursor changes are validated before commit—~27s microbuilds and 3–5× lower token burn on retries in their tests.
An HN commenter described a Claude Code setup with reusable skills for a UI library/Storybooks/APIs, a post-change review agent, PR checklist reviews, scheduled docs/coverage agents, and ticket triage that leaves a halfway-done plan on each issue.
Kelos orchestrates Claude Code, Codex, Gemini, or OpenCode in ephemeral Kubernetes pods as CRDs—watching bug issues, drafting fix PRs, self-reviewing, and retrying when CI fails—while the author still does final review and merge.
Room Service 3.3’s local MCP gives coding agents disk/port/Git visibility and actions like stop-port or trash build data—off by default, local socket only, permanent actions need in-app approval, every action logged.
