AI AGENT SECURITY & BEHAVIORAL GOVERNANCE
The runtime security and active-defense platform for AI agents
THE FFAX PLATFORM
Four layers of defense, one decision engine
Runtime Control
Hooks and runtime injection observe and control every agent action before it runs.
Network Gateway
Zero egress by default, deep packet inspection and data-smuggling detection.
Semantic Review
AST rules and a local security model judge what AI-written code really does.
Active Deception
Decoys, a read-only maze and honey-agents turn a hijack attempt into intelligence.
Threat Response
Real-time analysis with recommended or automated countermeasures.
Zero-Trust Workspace
MicroVMs, an air-gapped network and taint tracking for high-risk teams.
ACTION LOG & REPLAY
Every action, with the full story behind it
Which file it read, which command it ran, which domain it called: all logged locally. Every blocked action comes with its context, down to the input that led the agent there.
WORKS ACROSS AGENTS
One policy layer for Claude Code, Cursor, Codex and MCP
DESIGN PARTNER PROGRAM
Join as a design partner and shape the product with us
Design partner
Free
during the 6-week demo
Runtime security for AI coding agents, built for mid-size engineering teams. Local-first, with an open-source client.
Open-source client that runs locally
Action log that never leaves your machine
Three default policies out of the box
Decoy layer: fake credentials and honeypot tools
One-page console with live actions and alerts
Self-hosted deployment for teams
WHY FFAX
Built for security teams, invisible to developers
Local-first
Logs and policies stay on the developer’s machine. Code and prompts never leave.
Five-minute setup
One command detects and connects your agents, with default policies ready to go.
Self-hosted
The team edition runs inside your network, with no third-party cloud in the path.
Open-source client
Auditable code, built in the open on GitHub.
Cross-vendor
One event format and one policy set for Claude Code, Cursor and Codex.
Deterministic alerts
Decoys fire on touch, not on a model’s guess. Zero false positives.
ROADMAP
The first six weeks: a live demo of a blocked attack
FREQUENTLY ASKED QUESTIONS
What teams ask before getting started