Notes from building openma. Agent design, integrations, billing, the occasional postmortem.
Practical migration guide. What stays the same, what changes, and the exact steps to switch your client code, sessions, vaults, and integrations to the open-source platform.
How Open Managed Agents is structured under the hood. Durable Objects with embedded SQLite as the per-session log, Cloudflare Containers as the sandbox, R2 for blobs, and the deliberate split between the brain (harness) and the body (sandbox).
End-to-end guide to deploying Open Managed Agents on Cloudflare. Workers + Durable Objects + Containers + R2. Wrangler config, secrets, custom domains, and what to do when each piece breaks.
What's actually shipping in 2026 if you want an open-source alternative to Anthropic's Managed Agents. Honest comparison: Open Managed Agents, LangGraph, AutoGen, CrewAI, plus what's still missing.
Workers + Containers + D1 + R2 = an agent platform with very little ops. The substitution guide for anyone who can't be on Cloudflare.
Side-by-side technical comparison of Anthropic's Managed Agents and the open-source Open Managed Agents project. API surface, runtime model, sandbox, billing, and where each one actually fits.
Why we're building an open-source meta-harness for AI agents — and why BYOK matters.
