{"slug": "agent-os-a-local-first-harness-around-coding-models", "title": "Agent OS: A Local-First Harness Around Coding Models", "summary": "A developer released Agent OS, an open-source local-first AI project operating system that provides a harness around coding models. The system separates responsibilities between a Main Agent and a Coding Agent, with sandboxed workspaces and explicit verification before external actions. The developer used Agent OS to build and deploy Pulseboard, a full-stack SaaS, to test its recovery capabilities.", "body_md": "A coding model can generate code.\n\nThat does not mean it can reliably finish software work.\n\nBetween a plausible diff and a completed task, a real software agent still needs memory, execution boundaries, verification, recovery, permissions, and delivery infrastructure.\n\nThat is the problem I have been exploring through **Agent OS**, which I have now released as an open-source project.\n\n**GitHub:**\n\n[https://github.com/earthwalker17/agent-os](https://github.com/earthwalker17/agent-os)\n\nAgent OS is a local-first AI Project Operating System: a harness around coding models.\n\nThe core architecture separates two responsibilities.\n\nThe Main Agent handles:\n\nIt cannot edit repository code or execute shell commands.\n\nThe Coding Agent operates inside one sandboxed project workspace.\n\nIt can inspect files, edit code, and run bounded commands, but it cannot modify project memory or access another project’s workspace.\n\nThe two sides communicate through summaries and structured artifacts rather than unrestricted shared control.\n\nAgent OS does not accept the model’s own claim that a task is complete.\n\nA coding run must pass a real build or test command.\n\nIt can then:\n\nRecovery is deliberately limited. The goal is not unlimited autonomy, but controlled progress with evidence and an audit trail.\n\nGit pushes, pull requests, deployments, database migrations, and Stripe test-mode operations use explicit preview-and-confirm contracts.\n\nThe agent can prepare the operation, but it cannot silently mutate an external system because it inferred that the user probably wanted it.\n\nTo pressure-test the architecture, I used Agent OS itself to build and deploy **Pulseboard**, a full-stack SaaS, from an empty repository.\n\nThe process included real build, runtime, browser, visual, deployment, and database failures. Those failures became the test for whether the system could collect evidence, expose the problem, and recover instead of simply producing a confident success message.\n\nThis is an early public release.\n\nThe Windows setup is currently the most thoroughly tested. macOS and Linux users need to follow the manual setup instructions. At least one supported model provider API key is required.\n\nLocal-first means that project memory, workspaces, credentials, and execution records remain under the user’s control. It does not currently mean that every supported model runs locally.\n\nI am especially interested in feedback on:\n\nThe repository, architecture documentation, installation guide, and production showcase are available here:", "url": "https://wpnews.pro/news/agent-os-a-local-first-harness-around-coding-models", "canonical_source": "https://dev.to/_2e39841ea0f3747512e67/agent-os-a-local-first-harness-around-coding-models-238b", "published_at": "2026-07-10 16:48:13+00:00", "updated_at": "2026-07-10 17:15:35.481639+00:00", "lang": "en", "topics": ["ai-agents", "developer-tools", "ai-infrastructure", "ai-products"], "entities": ["Agent OS", "Pulseboard", "GitHub", "Stripe"], "alternates": {"html": "https://wpnews.pro/news/agent-os-a-local-first-harness-around-coding-models", "markdown": "https://wpnews.pro/news/agent-os-a-local-first-harness-around-coding-models.md", "text": "https://wpnews.pro/news/agent-os-a-local-first-harness-around-coding-models.txt", "jsonld": "https://wpnews.pro/news/agent-os-a-local-first-harness-around-coding-models.jsonld"}}