n0tduck1e/theharbinger

Harbinger is a security platform for non-human identities: AI agents, bots, and service accounts. It gives every agent a cryptographic identity, checks each request it makes against a policy before allowing it through, and pulls real secrets from your own vault only at the last moment. The agent itself never holds a standing credential. The project frames itself around non-human identity security in general, with AI agents as the sharpest example. Most current agent security tools either require rewriting your agent against an SDK, show you a dashboard of what credentials exist without changing how they work, or just move responsibility from one API key to another. Harbinger takes a different approach. It is a transparent proxy, so you point your agent's traffic at the gateway and it is covered. No code changes are needed, and third-party agent frameworks you do not control work too. Every request with credentials gets checked against two questions: is the agent allowed to access this website, and is the agent allowed to use this credential against this website? Only when both are true does the agent get to use the credential. Agents receive a placeholder instead of a real secret. The gateway swaps in the real value only after policy authorizes that agent for that target and confirms the request is actually headed to the registered destination. This stops a compromised agent from exfiltrating a secret to an attacker-controlled host. The system has several components. An admin server acts as the control plane, owning the database, policy storage, and enrollment review. A gateway server is the data plane that all agent traffic passes through. A command-line client handles setup and management. A browser UI covers everything except cryptographic operations. An optional wrapper daemon runs an existing agent as a child process and routes its traffic through the gateway automatically. The full README is longer than what was shown.