Agentmetry
Agentmetry provides an open-source, local-first flight recorder for AI coding agents such as Cursor, Claude Code, and Codex. It records every tool call, tags sequences with MITRE ATT&CK framework mappings, and correlates events into detections including data injection attacks. The tool runs entirely on the user's machine with no cloud dependency, making it suitable for platform teams, security engineers, and developers seeking transparent observability into agent behavior.
Open SourceApache-2.0CLISelf-hosted
8stars
2forks
5features
2026since
Overview
5 featuresAgentmetry sits in PulseGate's LLM eval & observability category. It focuses on detecting and auditing potentially malicious or unintended actions performed by AI coding agents in development environments. It is built as an open-source project for security engineers and platform teams. The project is open source (Apache-2.0). Agentmetry is available on the command line, and it can be self-hosted.
blitzcrieg1 builds and maintains Agentmetry, and it first shipped in 2026. The project is developed in the open on GitHub with 232 commits in the last 90 days. Key capabilities include ATT&CK Tagging, Session Recording, and Threat Detection. It exposes integrations via an MCP server.
Summary written by a language model from the project’s public pages.
- ATT&CK Tagging
- Session Recording
- Threat Detection
- Local-first
- MITRE Mapping
Tags
agent-securityaudit-trailmitre-attackai-agent-monitoringlocal-observability
AI capabilities
Code
Inference: Local
Built with & integrations
Hosting
Vercel
AI providers
anthropic
Connectors
MCP
Runs on
CLISelf-hosted
Detected from
- Vercel
- x-vercel-id header · x-vercel-cache header
Trust & compliance
License
Apache-2.0
Verified signals
Indexing history
2What PulseGate has recorded for this listing
- Indexed2 Aug · 18:13 UTCI pointed my agent security tool at myself and four of the bugs were mine verified against its public sourceSource: PulseGate · Open
- Indexed1 Aug · 17:36 UTCShow HN: Agentmetry – local-first flight recorder for AI coding agents verified against its public sourceSource: PulseGate · Open
Frequently asked questions about Agentmetry
- What does Agentmetry do?
- Agentmetry focuses on detecting and auditing potentially malicious or unintended actions performed by AI coding agents in development environments. It is catalogued under LLM eval & observability on PulseGate.
- Who should use Agentmetry?
- Agentmetry is an open-source project built for security engineers and platform teams.
- Does Agentmetry have a free plan?
- Yes — Agentmetry is open source under the Apache-2.0 license and free to use.
- What platforms does Agentmetry run on?
- Agentmetry runs on the command line. It can also be self-hosted.
- Is Agentmetry still maintained?
- The GitHub repository shows 232 commits in the last 90 days.
- Who develops Agentmetry?
- Agentmetry is developed by blitzcrieg1.
- When did Agentmetry launch?
- Agentmetry first shipped in 2026.
- Is Agentmetry open source?
- Yes — Agentmetry is open source under the Apache-2.0 license, developed on GitHub.
Similar apps
Closest matches by what these projects do