Overview
AmbiOS AI is a WebMCP-native collaboration and operations workspace for teams that need AI assistance without surrendering control. It brings context, incident response, documentation proposals, guardrails, approvals, budgets, and integrations behind shared server-side policy.
The current release catalogs 29 tools and mounts an explicitly safe read-only subset in compatible browsers. Write-capable actions remain gated until their provider adapters, approval paths, verification evidence, and external side effects are proven.
What I built
- Operational context and incident-response surfaces
- Guardrails and approvals around governed workflows
- Budget and integration controls for agent inspection
- Documentation proposals that remain reviewable before publication
- WebMCP tools behind shared server-side authentication and governance
Stack
Next.js, Hono, TypeScript, WebMCP, Cloudflare Workers, D1, KV, R2, Queues, Supabase Auth, and policy-aware server-side adapters.
What I learned
Agent products need an operating model, not only a chat interface. Context, permissions, budgets, approval states, and audit evidence are product primitives when an agent can inspect or affect real systems.
