We just shipped govern() SDK, a single method that sits between your app and Anthropic's Managed Agent runtime. Every tool call (bash, file writes, web fetches) flows through your org's policies before executing.

It evaluates each tool call against your workspace policies. Low-risk tools auto-approve instantly. High-risk tools escalate to humans via alerts and dashboard. Everything lands in a tamper-proof, ECDSA-signed audit trail.

It is very easy to make any agent govern()-able, here's how:

  const al = new AgentLattice({ apiKey, anthropicApiKey });
  const result = await al.govern(agentId, "Analyze this report");

X post: https://x.com/gadbadd/status/2046016210438091196

Blog: https://agentlattice.io/blog/govern-anthropic-managed-agents

Docs: https://agentlattice.io/docs/managed-agents

GitHub: https://github.com/AgentLattice/govern-anthropic-demo