numbat

What is numbat?

numbat is an open-source command-line tool, published under the perplexityai GitHub organization, that gives security teams visibility into what AI agents do on a machine. It watches the desktop, CLI, IDE and gateway agents it supports, using local hooks, plugins, OTLP/HTTP logs and the session files agents leave on disk. Everything is normalized into one event model and checked by a CEL rule engine with built-in rules, multi-step sequence rules and custom YAML rules, and detection runs on the endpoint itself. Blocking is off by default and works only through supported pre-action hooks for rules you mark to enforce. It can also reconstruct past activity from on-disk sessions and package a case bundle for an investigation. numbat is free under the Apache License 2.0 and ships as a single binary for macOS, Linux and Windows.

Use cases

• Monitor AI coding agent activity on endpoints
• Detect secret access followed by data upload
• Block risky agent actions with enforce rules
• Reconstruct past agent sessions for forensics
• Write custom detection rules in YAML
• Build case bundles for incident investigation
• Inventory which AI agents run on a machine

Who it's for

• Security engineers
• Incident responders
• Platform teams
• Developers running AI agents

Why use numbat

• Runs detection on the endpoint, so records stay local unless you choose to send them to an HTTP sink.
• Starts in monitor-only mode, which lets teams review findings before turning on any blocking.
• Covers live activity and past sessions with one rule engine, so the same rules apply to both.
• Custom rules are plain YAML and can be tested with a built-in command before rollout.
• Free and open source under Apache 2.0, with checksums published for each release.

numbat features

• Live monitoring through hooks, plugins and OTLP/HTTP logs
• Local detection with built-in CEL rules
• Multi-step sequence rules and custom YAML rules
• Opt-in blocking via supported pre-action hooks
• Forensic reconstruction from on-disk session artifacts
• Versioned NDJSON records with published JSON Schemas
• Read-only artifact scanning with secret redaction
• Per-session timelines and agent discovery
• Portable case bundles with SHA-256 manifests
• Single binary built without cgo

numbat integrations

• Codex
• Claude Code
• OpenClaw

How to get started with numbat

1. Install a release binary for your OS, or run go install with Go 1.26.9 or newer.
2. Run numbat agents and numbat scan to find agents and scan their on-disk sessions without changing any configuration.
3. Run numbat hook install for a supported agent to start live monitoring in monitor-only mode.
4. Review findings in the NDJSON record stream, written to stdout, a local file or an HTTP sink.
5. Optionally copy a shipped rule into your own policy folder, mark it enforce, and install it for pre-action blocking.

numbat limitations

• Blocking is off by default and limited to supported pre-action hooks
• All shipped rules are monitor-only until you copy and mark them to enforce
• Findings are rule matches, not proof of compromise
• Reconstruction cannot recover activity an agent never saved

numbat pricing

numbat is open source, so you can use and self-host it for free.

Prices can change and may vary by country or exclude tax. Check the official pricing page.

numbat alternatives

The main numbat alternatives listed on Debuly are E2B, Claude Agent SDK, LangChain Deep Agents, Letta, Context Data and Fireraven. The table compares pricing and platforms; order is A–Z.

1. 

   Claude Agent SDK

   Anthropic's Python SDK for building agents on the Claude Code runtime, with built-in tools, custom tools, hooks and MCP servers.

   Developer platformsAutomation and agentsCoding
   Visit Claude Agent SDK (official site)
2. 

   Context Data

   InterLock proxy that governs what AI agents can read and write in databases, files and APIs, with redaction and approvals.

   Developer platformsAutomation and agents
   Visit Context Data (official site)
3. 

   E2B

   Cloud sandboxes where AI agents run code safely, with Python and JavaScript SDKs and per-second usage billing.

   Developer platformsAutomation and agentsCoding
   FreemiumVisit E2B (official site)
4. 

   Fireraven

   Runtime guardrails and automated red-teaming that secure AI agents and help prove compliance.

   Developer platformsAutomation and agents
   Visit Fireraven (official site)
5. 

   LangChain Deep Agents

   Open-source LangChain agent harness: long-term memory, a virtual filesystem, subagents and planning included

   Developer platformsAutomation and agentsCoding
   Open sourceVisit LangChain Deep Agents (official site)
6. 

   Letta

   Platform, formerly MemGPT, for building stateful AI agents with persistent memory, from a terminal app, SDK or API.

   Developer platformsAutomation and agentsCoding
   FreemiumVisit Letta (official site)

numbat FAQ

Sources & verification

Facts on this page come from the official sources below, each with the date we read it. Written by Debuly; the vendor did not review this page.

Image credits

• Logo: from avatars.githubusercontent.com · fetched 24 Oct 2026 · Vendor trademark, used only to identify the product; removed on request
• Cover: none collected yet, so a generated tile is shown.

Spot an error? Report an issue with this listing. We also remove a logo or image on the owner's request.