Holmes: Open-Sourcing a Model-Agnostic Cloud Coding Agent

Episode 1 · Developer Open-Sources holmes, a Cloud Agent Coding Tool with Swappable Models and Isolated Environments (2026-10-09, 12 posts)

On October 9, developer Lucas Meijer open-sourced and released for free holmes, a cloud agent coding tool he had been using for months. The tool aims to address three pain points in his AI-assisted programming: not being locked into any single AI lab, cloud agents running in isolated environments unable to touch his local laptop, and staying on top of codebase state without hand-writing code.

Confirmed

• holmes runs cloud agents on the user's own hardware, assigning each task an isolated remote development environment with its own browser, terminal, and VS Code instance.
• The tool takes an "any agent, any model, any subscription" approach — for example, running the original Claude Code with the user's existing Claude subscription.
• The project is free and open source, and the mobile experience is highlighted by the author as a key feature.
• holmes is specifically designed for evaluating agents' work.

Why it matters

• It offers a free, open-source option for a cloud-agent workflow that isn't locked to a single lab and uses your own subscriptions — appealing to developers who want to keep existing subscriptions (like a Claude subscription) while running agents in isolated environments; the per-task isolated environments also reduce the risk of agent mistakes affecting the local machine.

• Dev ships free open-source cloud coding agent tool with no single-lab lock-in — lucasmeijer · 2026-10-09
• Lucas Meijer open-sources cloud agent coding tool with per-task remote dev environments — lucasmeijer · 2026-10-09
• Open-source holmes: per-task cloud agent workspaces with browser, terminals, VS Code — lucasmeijer · 2026-10-09
• Remote dev setup: each workspace gets its own agent, browser, and terminals — lucasmeijer · 2026-10-09
• Dev shows remote workspaces running any agent on any model, incl. vanilla Claude Code — lucasmeijer · 2026-10-09
• Agents now open the exact app screen where their change lands, cutting eval legwork — lucasmeijer · 2026-10-09
• Dev shares agent workflow: background workspaces protect your focus — lucasmeijer · 2026-10-09
• Developer Now Does Half His Coding From His Phone Using AI Workspaces — lucasmeijer · 2026-10-09
• Keep secrets out of agent workspaces: proxy-swapped tokens for safe AI coding — lucasmeijer · 2026-10-09
• Where to run your agents: Mac for quick start, a server for the long run — lucasmeijer · 2026-10-09
• Dev tours his AI workstation: readable bash, parked workspaces, usage rings — lucasmeijer · 2026-10-09
• Open-source AgentsInTheCloud runs coding agents on a server; GrokBot auto-built agents from X replies — lucasmeijer · 2026-10-09

Episode 2 · Developers Share Multi-Machine Agent Workflows with SSH and Mac Simulators (2026-10-09, 3 posts)

Developer lucasmeijer shared his multi-machine agent workflow on AgentsInTheCloud, giving agents SSH keys to schedule other machines and an external Mac for iOS simulator testing. The platform's author explained that agents learn their environment via fully transparent, inspectable prompts.

• AgentsInTheCloud author: agent environment prompt is fully visible via system prompt button — lucasmeijer · 2026-10-09
• Running agents across machines: give the agent SSH keys to other machines it can use at will — lucasmeijer · 2026-10-09
• Attach an external Mac so your agents can drive the iOS simulator — lucasmeijer · 2026-10-09