What is Omnodex?

Omnodex is a security and observability platform that makes every connection, tool call, credential use, and data flow in AI agent execution traceable and auditable.

The Problem

AI agents execute autonomously - calling tools, reading files, using credentials, reaching third-party APIs - with no persistent record of what happened or why. Security and compliance teams have no way to audit what an agent did, what data it touched, or whether it behaved within policy.

Omnodex solves this by intercepting agent activity at the execution layer, tracing every action, analyzing it for risk, and surfacing results for human review.

How It Works

The core pipeline is:

1. Intercept - Hook into AI agent execution (Claude Code, OpenAI Codex, Cowork Desktop) with zero overhead to the agent itself
2. Trace - Capture every tool call, file access, credential use, and network connection as structured events
3. Analyze - Run configurable risk detection rules against the event stream in real time
4. Report - Surface findings in a dashboard with full event context for review

Key Properties

• Zero token overhead - Hooks fire asynchronously outside the agent’s context window. The agent never sees Omnodex.
• Non-blocking - Interception never slows down or interrupts the agent’s execution.
• Append-only audit log - Events are written to an immutable log that cannot be retroactively altered.
• Local-first - The free tier runs entirely on your machine. No data leaves your environment unless you opt into cloud features.

What Gets Captured

Omnodex traces a complete picture of agent activity:

• Tool invocations (what was called, with what parameters, what it returned)
• File system access (reads, writes, deletions)
• Credential usage (which API keys or tokens were used, and where)
• Network connections (outbound requests to third-party services)
• Session lifecycle (start, end, duration, models used)

Next Steps

Ready to get started? Head to the Quick Start guide to have Omnodex running in under five minutes.