Devs debate multi-machine agent networks: trust and policy, not messaging, is the hard part
No_Praline7219 · reddit · 2026-09-06
A Reddit developer wants his Claude Code session on a MacBook to delegate tasks to a Mac mini over Tailscale—executing, asking mid-task questions, and reporting back—with a friend's machine joining under a restricted role like reviewer or GPU worker.
His key takeaways:
- The hard part is trust, not messaging: directional policy (macbook→mini auto-approved within a verb allowlist, mini→macbook requires approval, default deny), with the friend's inbound rules taking precedence on his node.
- Isolation, per-node keys, revocation, and subscription billing remain open problems.
He surveyed existing tools: A2A has task lifecycle but no peer authorization; cmesh does policy-gated delegation over Tailscale but is very early; Agent Teams is single-machine by design. He asks whether to build from scratch or use git-as-queue, per-machine MCP servers, SSH allowlisting, or a broker.
More from coding & agent
- Spec-to-APK entirely on an Android phone via Termux CLI and AI — Digital_Otorongo · 2026-09-06
- Astra review: executes hour-long agentic tasks while iterating on the plan without losing the thread — ivan_bezdomny · 2026-09-06
- Stop picking coding models on vibes: run your own A/B/C refactor tests — robleclerc · 2026-09-06
- Orca, the 62k-star ADE for running parallel coding agents, recommended for Mac Mini dev setups — kieranklaassen · 2026-09-06
- Dev Abandons LLM Agent Harness After 883 Commits and 8 Months, Open-Sources the Wreckage — dancingwithlies · 2026-09-06
- Dev one-shots a UEVR fork with GPT 6 Astra for true stereoscopic VR windows — pvncher · 2026-09-06