Know Exactly What
Your AI Agents Did

Log every file read, file write, and command executed by your AI Agent.
Query it, filter it, replay it. No cloud. No guesswork.

Gryph Hero
Blind spot illustration
The Blind Spot

AI Agents Act,

You Find Out Later

Agents operate autonomously across your environment. They read files, execute commands, and modify code as part of normal workflows but there's no clear visibility or transparency.

Gryph record illustration
With Gryph

Complete Record of

Every Action

Gryph hooks into your AI agents and records every action they take be it reading files, writing code, executing commands, session activity, etc. All stored locally in a structured audit trail you can query anytime.

Local by design illustration
Local By Design

Data Stays On

Your Machine

Gryph has no cloud backend. No telemetry. No data ever leaves your machine. Configurable retention, logging levels, and fully inspectable detection rules as privacy shouldn't require trust.

Claude
Cursor
OpenAI
Windsurf
Gemini
How It Works

Simple Local Setup
No Workflow Change

Gryph works with your workflow. Simply install hooks, use your agent as normal, query the trail when you need it.

Install hooks illustration
[1]

Install Gryph Hooks

Install hooks for all detected agents using gryph install commands. Gryph attaches lightweight hooks to your AI agents.

Capture and store illustration
[2]

Capture & Store Every Action in SQLite

Every time the agent performs an action, the hook sends a JSON event to Gryph which stores it locally in a SQLite database.

Query and analyze illustration
[3]

Query, Replay, & Analyze

Once all the actions are stored, you can use CLI tools to inspect logs, replay sessions, or export data.

Complete Security

Built For Teams
To Be Accountable

Gryph gives you the visibility layer that AI coding agents were missing. See what ran, prove what changed, and debug what went wrong.

Audit trail log view

Audit Trail

Every session gets a unique ID and timestamp. Export to JSON or CSV for compliance, incident review, or team sharing.

Full transparency overview stats

Full Transparency

Every file read, every file written, every command run, is captured as a structured event and queryable by agent, action, or file path.

Session replay view

Session Replay & Debugging

Step through what the agent did, action by action, diff by diff. When something breaks, the full session is already on record.

2M+
Packages Scanned
100K+
Components Analysed
1000+
Projects Secured
80%
Threats Blocked

Monitor Code
AI Agents Ship

Every rule, every hook, fully inspectable. Install in seconds.