The Agent Governance Toolkit (AGT) provides policy enforcement, zero-trust identity, and audit logging for autonomous AI agents at the application middleware layer. It operates as a Python library that intercepts tool calls in deterministic code to enforce governance rules before any action is executed.
Project overview
By intercepting tool calls in deterministic code before they reach the network, the project structurally prevents denied actions rather than relying on prompt-level safety mechanisms.
Project type
AI Agent · Infrastructure
Use cases
Automation
Deployment
Refer to project documentation
License
MIT
Best for
Developers, AI engineers, and enterprise teams who need to attribute agent actions, verify permissions, and log decisions for autonomous workflows.
Key capabilities
Intercepts tool calls in deterministic code to enforce YAML, OPA, or Cedar policies before the model's intent reaches the wire, making denied actions structurally impossible.
Provides agent identity, routing, and trust mesh using SPIFFE, DID, and mTLS to attribute actions to specific agents.
Records every decision, including the active policy, agent requests, and allow or deny reasons, to provide proof for auditors and regulators.
Limitations and risks
Governance is enforced at the application middleware layer, not the OS kernel level. The policy engine and agents share the same process boundary.
Operating with third-party agent frameworks or services carries inherent risks regarding data retention and location that are outside this project's control.
Getting started
Install via a single command: pip install agent-governance-toolkit[full]. After installation, import the govern function from agentmesh.governance and wrap existing tools, for example: safe_tool = govern(my_tool, policy="policy.yaml").
Evidence and sources
GitHub project description: AI Agent Governance Toolkit — Policy enforcement, zero-trust identity, execution sandboxing, and reliability engineering for autonomous AI agents. Covers 10/10 OWASP Agentic Top 1…
README: Every tool call, message send, and delegation is intercepted in deterministic application code *before* the model's intent reaches the wire. Actions the AGT kernel denies are not…
README: Govern any tool function in two lines: ```python from agentmesh.governance import govern safe_tool = govern(my_tool, policy="policy.yaml") # every call checked, logged, enforced `…