# Fencio

> Fencio is the Zero Trust Firewall for AI Agents. It learns an agent's behavioral boundaries during pre-production evaluation, then enforces those constraints deterministically at runtime, outside the LLM itself, intercepting tool calls and state transitions before actions occur.

Fencio is built for engineering, security, and compliance teams deploying autonomous AI agents in production, especially in regulated industries like fintech, legaltech, healthcare, and cybersecurity. Founded by Aashish Bharadwaj and Avkash Kathiriya, based in San Francisco, backed by Entrepreneurs First (EF).

## Products

- [Prism](/products/prism): Runtime enforcement engine. Deterministic, rule-based interception of agent tool calls and state transitions, written in Rust.
- [Shark](/products/shark): Pre-production AI red teaming and attack surface assessment, feeding the behavioral constraints Prism enforces.

## Docs

- [Documentation Home](/docs): Overview of Prism's adaptive runtime enforcement model.
- [API Reference](/docs/implementation/api): HTTP API for enforcement, policy management, and monitoring.
- [CLI Reference](/docs/implementation/cli): Command-line tool for managing Prism services locally.
- [Docker Deployment](/docs/deployment/docker): Run the full Prism stack via Docker Compose.
- [Kubernetes Deployment](/docs/deployment/kubernetes): Deploy Prism on Kubernetes with the official Helm chart.

## Integrations

- [LangChain / LangGraph](/docs/integrations/langchain): Official Prism integration for LangChain and LangGraph agents.
- [Proxy](/docs/integrations/proxy): Enforce policies at the network level for any HTTP/HTTPS traffic, no SDK changes required.
- [MCP Server Wrapping](/docs): Wrap existing MCP servers to enforce tool-call policy without modifying the agent.

## Research

- [Blog](/blogs): Security research, tracked AI agent incidents, and case study analysis.

## Contact

- Website: https://fencio.dev
- Early Access: https://fencio.dev