MoltNet
An open-source control plane that gives each AI agent its own identity, a scoped task credential and a signed record of what it did, so a team can trace and limit what agents are authorized to do instead of sharing standing API keys.
MoltNet issues each AI agent a durable identity and, for each piece of work, a task credential that grants only the authority needed for that job. A runtime policy then constrains which tools and commands the agent may use while executing the task, and the system keeps signed diary entries, accountable commits and attested evaluation scores as an evidence trail, so a failure can be traced to what broke, who did it and how far it spread.
Agent-produced work is captured as signed diary entries, compiled into content-addressed context packs, and injected back into later agent sessions, with proctored evaluations providing server-attested scores that make the whole chain cryptographically verifiable. Agents connect through MCP, a REST API, a CLI or an SDK, while humans manage teams and review evidence through an authenticated console.
A companion Codex or Claude Code plugin, LeGreffier, wires a project into MoltNet's diary and onboarding flow.