Agentic Guardian
Delegant’s framework and tooling for securing and managing agent permissions—built on the open-source Thebe Protocol. Designed to make agentic solutions safe to run in real environments, from personal workflows to enterprise deployments.
Demo: www.delegant.online
What it does
Agentic Guardian focuses on permissioning and control—so agents can be powerful without being dangerous.
Typical controls
These are the kinds of enforcement points we help you design and implement around your agents and tools.
- Tool access control (allowlists/denylists)
- Scoped secrets access (per tool + environment)
- Human-in-the-loop approvals for high-impact actions
- Rate limits and time-based permissions
- Structured outputs and schema validation
- Redaction and sensitive-data handling
- Action logging and trace correlation
- Evals, regression testing, and safety checks
Try it & explore the source
Agentic Guardian is available to try online and as open source. The reference implementation uses zero-trust verification, policy enforcement, and tamper-evident audit logging—compatible with agent stacks like LangGraph, Agno, and OpenClaw.
Whether you’re using OpenClaw, NanoClaw, Claude Cowork, or your own internal tools, Agentic Guardian adds consistent permissioning and governance around tool-use and actions.