PAGES
← all tools
◆ AI Observability & Evals

numbat

v0.3.0 open-source

numbat provides endpoint visibility into AI agent activity, with local detection, optional pre-action blocking, and forensic reconstruction.

Summary

numbat is an open-source, single-binary command-line tool that watches AI agents running on a laptop or server and reconstructs what they did. It's free to use, distributed as a binary for macOS, Linux, and Windows, or installable via go install, with no server component required. It works through local hooks and plugins for live agents like Codex, OTLPOTLPOpenTelemetry Protocol, a standard way to send traces, metrics and logs between applications, collectors and monitoring systems. It lets telemetry move across compatible tools./HTTP log ingestion, and read-only scanning of on-disk session artifacts, normalizing everything into one event model that a local CELCELCommon Expression Language, an open-source expression language maintained by Google, used to embed lightweight, safe policy evaluation logic into tools without running a full scripting engine. rule engine evaluates for detection and, optionally, pre-action blocking. It's built for teams securing AI agent deployments — platform and detection engineers rather than end users — who need endpoint visibility into agent behavior rather than network- or cloud-layer monitoring. Findings and enforcement decisions are written as versioned NDJSONNDJSONA text format where each line is a self-contained JSON object, making it easy to stream, append, and parse large datasets record-by-record without loading the whole file into memory., with case bundles for investigation.

numbat provides endpoint visibility into AI agent activity, with local detection, optional pre-action blocking, and forensic reconstruction.

How numbat helps

Which AI agents are actually running on this machine right now?

reports discovered agents by name with their config signals, on-disk artifacts, live capture mode, and whether a hook is already wired, before anything is installed

What did an agent actually do during a past session?

reconstructs a chronological per-session timeline from on-disk session artifacts alone, with no prior instrumentation required

Did an agent try something that matches a known risk pattern?

evaluates normalized events against local CEL rules, including multi-step sequence rules, and writes findings with the source event references that triggered them

Would a custom policy have actually caught the behavior I'm worried about?

validates and test-runs operator-authored YAML rules against a saved NDJSON fixture, showing what the rule would have flagged before it's ever loaded live

Can I stop an agent from writing to a sensitive file before it happens?

supports opt-in pre-action blocking on supported synchronous hooks, but only for rules explicitly marked to enforce — every shipped rule stays monitor-only until an operator opts it in

Can I hand this incident to someone else without exposing the whole transcript?

builds a portable case bundle with a SHA-256 manifest whose records redact secrets and omit raw transcripts by default, keeping raw evidence out unless explicitly added

all 13 features, with the evidence for each →

Languages

Not documented

Features

13 capabilities across 3 capability groups · 5 backed by code, an API document or a real run

Built from numbat's releases, docs, README and our own runs. “Seen” marks the month we first spotted a release the vendor didn't date. Group names and tags are our labels.

Compare with peers →

  • numbat covers 1 of the 6 ai evaluation and governance capabilities we track.
  • numbat covers 2 of the 11 logs and telemetry capabilities we track.

What numbat offers you

  • Command-line interface8
  • JSON data export3
  • User-defined detection rules2
LLM application tracing 2 features AI evaluation and governance since 2026
Session timeline reconstruction shipped Organises collected or scanned artifact data into a chronological, per-session view of agent activity. 3 releases · first seen Jul 2026

Shared capability: LLM application tracing

Also counts toward: Command-line interface

release

  • Improves subagent attribution in timeline and event records, preserving subagent identity and correlation across capture paths. v0.3.0 · Oct 2026 · source · release history
  • Adds numbat timeline [--agent NAME ... | --path FILE|DIR ...] to reconstruct per-session chronological views from agent artifacts. v0.1.1 · Jul 2026 · source · release history
  • Adds forensic reconstruction of AI agent sessions v0.1.1 · Jul 2026 · source · release history
Live monitoring with event streaming shipped Runs live hooks against a running agent and writes all observed events, findings, and enforcement decisions to a local record file as they happen. 2 releases · 1 other source · first seen Jul 2026

Shared capability: LLM application tracing

release

example

  • Install live monitoring for Codex and stream all events, findings, indicators, and enforcement decisions to the default record file. numbat hook install --agent codex --emit all v0.1.1 · Jul 2026 · source
Log collection 1 feature Logs and telemetry since 2026
OTLP log collection shipped Listens for OTLP log payloads sent over the network from supported agents and converts them into normalised records. 1 release · first seen Jul 2026

Shared capability: Log collection

Also counts toward: Command-line interface, JSON data export

release

  • Adds numbat collect [--addr HOST:PORT] to receive OTLP/HTTP protobuf logs from supported agents and emit normalized records. v0.1.1 · Jul 2026 · source · release history
Security event correlation 1 feature Logs and telemetry since 2026
Detection rule management verified Evaluates built-in and custom YAML detection rules against agent events locally, and provides tooling to author, validate, and test operator rules. 4 releases · 3 other sources · first seen Jul 2026

Shared capability: Security event correlation

Also counts toward: Command-line interface, User-defined detection rules

release

  • Adds rules test --json flag to numbat rules test for machine-readable rule-test output suitable for automation pipelines.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠ v0.3.0 · Oct 2026 · source · release history
  • Adds numbat rules check, numbat rules list, and numbat rules test --fixture FILE for validating, compiling, and evaluating CEL and multi-step sequence rules against NDJSON event fixtures. v0.1.1 · Jul 2026 · source · release history
  • Adds --rules-dir DIR (repeatable) and --no-builtin-rules flags to scan, collect, hook install, and rules check|list|test for loading operator-authored YAML rules or replacing built-in rules by id. v0.1.1 · Jul 2026 · source · release history
  • Ships built-in CEL and multi-step sequence detection rules evaluated locally; supports custom YAML rules in operator-controlled directories. v0.1.1 · Jul 2026 · source · release history

command line

example

  • Validate custom operator rules directory numbat rules check --rules-dir ./numbat-policy
Other / not yet classified 9 features Features we have not matched to a capability yet. A suggested match is marked.
Live agent hook management shipped Installs, removes, and inspects live integrations that let numbat intercept activity from supported AI agents. 4 releases · 2 other sources · first seen Jul 2026

Also counts toward: Command-line interface

release

  • Adds Muse Code live-hook monitoring, opt-in enforcement, and forensic at-rest session scanning via numbat hook install --agent muse-code and numbat scan --agent muse-code. v0.3.0 · Oct 2026 · source · release history
  • Rerun numbat hook install for each installed agent with the same options to update generated hooks and plugins after upgrading. v0.3.0 · Oct 2026 · source · release history
  • Adds numbat hook install --agent NAME|all and numbat hook uninstall --agent NAME|all to wire and remove live integrations for supported desktop, CLI, IDE, and gateway AI agents. v0.1.1 · Jul 2026 · source · release history
  • Adds numbat hook status [--agent NAME|all] to inspect numbat-owned live integration configuration. v0.1.1 · Jul 2026 · source · release history

example

  • Check hook status for Codex numbat hook status --agent codex
  • Install live monitoring hook for Codex numbat hook install --agent codex --emit all
NDJSON record emission verified Outputs all events, findings, indicators, and summaries as versioned NDJSON records with published schemas. 4 releases · 1 other source · first seen Jul 2026

Also counts toward: JSON data export

release

  • Output schema is now 0.5.0; update any downstream validators that enforce an exact version or closed schema. v0.3.0 · Oct 2026 · source · release history
  • Retains tool inputs and results in the event record so CEL rules and content exports can act on them, including MCP tool inputs and browser context. v0.3.0 · Oct 2026 · source · release history
  • Emits versioned NDJSON records (schema v0.2.0) for events, findings, enforcement decisions, indicators, and scan summaries, with JSON Schemas published under docs/schema/v0.2.0/. v0.1.1 · Jul 2026 · source · release history
  • Adds --emit all flag to hook install to write events, findings, indicators, and enforcement decisions to ~/.numbat/records.ndjson. v0.1.1 · Jul 2026 · source · release history

command line

  • --json — emit a machine-readable NDJSON result stream v0.3.0 · Oct 2026 · command-line history
Pre-action enforcement blocking shipped Blocks an agent action at the point of interception when a matching rule has enforcement enabled, with blocking off by default. 3 releases · 2 other sources · first seen Jul 2026

Also counts toward: User-defined detection rules

release

  • Adds opt-in enforcement for statically resolvable POSIX compound commands via rules marked enforce: true, extending blocking coverage beyond single-command pre-action hooks. v0.3.0 · Oct 2026 · source · release history
  • Adds opt-in pre-action blocking via --enforce on hook install; rules with enforce: true in an operator --rules-dir emit enforcement decisions; all shipped rules are monitor-only by default. v0.1.1 · Jul 2026 · source · release history
  • Supports optional pre-action blocking of AI agent requests v0.1.1 · Jul 2026 · source · release history

docs

  • Monitors AI agent activity on endpoints with local CEL-based detection, optional pre-action blocking, and forensic reconstruction from session artifacts.

example

  • Enforce a custom operator policy that blocks SSH authorized_keys writes by Codex — copy the shipped rule, add enforce: true, validate, then reinstall with enforcement enabled. numbat rules check --rules-dir ./numbat-policy && numbat hook install --agent codex --emit all --rules-dir ./numbat-policy --enforce v0.1.1 · Jul 2026 · source
AI agent discovery shipped Scans the endpoint to identify which supported AI agents are installed or active, along with their configuration state. 1 release · 4 other sources · first seen Jul 2026

Also counts toward: Command-line interface

release

  • Adds numbat agents [--all] for read-only discovery of supported local agents, reporting config signals, artifacts, live capture mode, and wired status. v0.1.1 · Jul 2026 · source · release history

docs

  • Monitors AI agent activity on endpoints with local CEL-based detection, optional pre-action blocking, and forensic reconstruction from session artifacts.

example

  • Discover which AI agents are present on the endpoint before installing any hooks. numbat agents --all v0.1.1 · Jul 2026 · source
  • Scan only Codex agent numbat scan --agent codex
  • List discovered AI agents numbat agents
Bounded message body inspection verified Inspects agent session content within defined limits on size and byte ranges. 2 releases · 2 other sources · first seen Aug 2026

Also counts toward: Command-line interface

release

  • Adds bounded message body inspection, enabling content analysis within defined limits on message size. v0.2.0 · Aug 2026 · source · release history
  • Supports bounded message body inspection, limiting content analysis to defined byte ranges. v0.2.0 · Aug 2026 · source · release history

command line

Forensic artifact scanning shipped Reads on-disk agent artifacts in a read-only pass and reconstructs activity records without requiring any prior setup or instrumentation. 2 releases · 2 other sources · first seen Jul 2026

Also counts toward: Command-line interface, JSON data export

release

  • Adds numbat scan [--agent NAME ... | --path FILE|DIR ...] for read-only forensic reconstruction from on-disk agent session artifacts, emitting versioned NDJSON records without prior instrumentation. v0.1.1 · Jul 2026 · source · release history
  • Supports read-only artifact scanning with secret redaction; raw evidence files are opt-in via case bundles. v0.1.1 · Jul 2026 · source · release history

example

  • Forensically scan saved Muse Code sessions without prior instrumentation — useful for incident review on a machine where numbat was not previously installed. numbat scan --agent muse-code v0.3.0 · Oct 2026 · source
  • Scan all discovered parser-backed agents numbat scan
Portable case bundles shipped Packages records into a self-contained, integrity-manifested bundle that can be verified and optionally includes raw evidence files. 4 releases · first seen Jul 2026

Also counts toward: Command-line interface

release

  • Adds --content full --content-scope messages flags to scope full-content exports to message bodies only, now that --content full also captures redacted tool arguments and results. v0.3.0 · Oct 2026 · source · release history
  • --content full now includes redacted tool arguments and results (source code, file contents, patches); use --content full --content-scope messages to restore message-body-only export behaviour. v0.3.0 · Oct 2026 · source · release history
  • Supports read-only artifact scanning with secret redaction; raw evidence files are opt-in via case bundles. v0.1.1 · Jul 2026 · source · release history
  • Adds numbat case build <case-id> --from FILE and numbat case verify DIR to create and verify portable case bundles with SHA-256 manifests. v0.1.1 · Jul 2026 · source · release history
Message output controls verified suggested Governs what content, including assistant and reasoning text, is emitted from inspected sessions. 3 releases · 1 other source · first seen Aug 2026

Suggested capability: Model output filtering. This is our guess, not confirmed.

release

  • Adds explicit message output controls to govern what content is emitted from inspected sessions. v0.2.0 · Aug 2026 · source · release history
  • Captures exposed assistant and reasoning text from AI agent sessions for analysis. v0.2.0 · Aug 2026 · source · release history
  • Adds explicit message output controls for governing what content is emitted from inspected message bodies. v0.2.0 · Aug 2026 · source · release history

command line

  • --include-reasoning — include source-recorded reasoning events when an integration exposes them v0.2.0 · Aug 2026 · command-line history
Remote record delivery verified suggested Continuously tails the local NDJSON record file and streams new records to a remote HTTP endpoint. 1 release · 1 other source · first seen Jul 2026

Suggested capability: Log collection. This is our guess, not confirmed.

release

  • Adds numbat ship --input-file F --http-url U to tail a local NDJSON record file to an HTTP endpoint for remote delivery. v0.1.1 · Jul 2026 · source · release history

command line

  • --max-batch-bytes — maximum uncompressed bytes per batch (1..4194304; v0.3.0 · Oct 2026 · command-line history

Lines in monospace are the tool's own words: help text parsed from its source, or an endpoint from its API document. Under release, a line is our summary of a dated release. Under docs, it is quoted from the tool's own documentation or website. Each links back to the source it came from.

Score card

  1. Shipping velocity releases a month, last 12 months
  2. Documentation quality release-notes rubric, 0–100
  3. Breaking-change discipline how rarely a release breaks users, and notice given
  4. Maintenance risk quiet time against its own cadence, maintainers
  5. Innovation & surface new capabilities, API and MCP surface
  6. Community health
  7. Contributor trend distinct committers, this year against last
  8. Maintainer concentration bus factor now against a year ago
  9. Issue backlog trend issues closed per issue opened, 12 months
  10. Responsiveness median time to a maintainer’s first reply

Nine scores for numbat, each built from the releases, notes and API history this site tracks — and, on Researcher and Business, where it stands among every measured tool in its category and how that has moved.

Score cards are part of every paid plan.

See plans →

Each score is 0–100 on a published rubric, derived from data this site already shows. A dimension we cannot measure says so — it is never scored 0 and never enters a rank.

How this is scored
  • Shipping velocity — releases a month over the last 12 months, on a log scale where 16 a month scores 100. Needs 3 releases in the window.
  • Documentation quality — the release-notes rubric published on the analytics page, unchanged.
  • Breaking-change discipline — the share of releases carrying a breaking change (none scores 100, half or more scores 0). Only projects that label breaking changes are measured: a clean record cannot be told from an unlabelled one.
  • Maintenance risk — quiet time measured against the tool's own release cadence (quiet longer than 90 days and three typical gaps scores 0), plus the number of active maintainers and whether contributors are growing, where known. Higher is lower risk.
  • Innovation & surface — new capabilities in the feature ledger this year, API and MCP operations added, and the share of commits that are features: the mean of whichever we hold.
  • Community health (open-source tools only) — each is a direction, not a size: contributors this year against last, the bus factor now against a year ago, and issues closed per issue opened, where steady scores 50, doubled 100 and halved 0; plus the median time to a maintainer's first reply (an hour scores about 100, a month or more 0). A commercial card has no repository and carries none of the four.
  • Documented delivery (commercial tools) — when a vendor's documented new capabilities have slowed against its own baseline, that counts against the overall standing as heavily as two low dimensions. Steady delivery, or too little history to judge, leaves the standing alone; it is never a sixth score.
  • Open-source tools are scored on repository signals; commercial tools on their published changelog. Ranks compare tools within one category and one profile and need at least five measured peers; the overall standing needs three ranked dimensions. Best-in-breed or consolidator is a shape, not a quality, and is never scored.

Release-note detail over time

See how much detail numbat puts in its release notes, quarter by quarter, and compare it with tracked tools in the same category. This history uses up to ten retained, reviewed notes per tool; sparse or blank older quarters may reflect that limit.

Checking release-note history…

Explore the full chart →

Coverage

Missing something? Tell us

What numbat connects to, runs on and detects, read from its release notes and docs. So far: 1 new integration.

Show

1 change, tracked since Oct 19, 2026.

  • Oct 192026

    new Muse Code live-hook and at-rest session support

    **Muse Code support:** Live monitoring, opt-in enforcement, and forensic scanning of saved sessions (#59).

    v0.3.0 Source ↗

How we count

Read from 1 of 3 release notes and docs updates we hold for numbat. "First seen" is the earliest note that names something; "last" is the latest. Our history may start partway through numbat's life, so an old integration can show a recent first-seen date.

Each change counts once: "added 40 checks" is one line, so the numbers measure how often coverage moves, not how big the catalogue is or how well it detects.

Releases

Show
Coverage in all entries

3 entries

  1. October 20261
  2. August 20261
  3. July 20261
    Show
    Coverage in all entries

    3 entries

    Project health

    Issue, contributor and maintainer activity in perplexityai/numbat, 2023-10 – 2026-09, as of 2026-09-25. Twelve-month figures first; the trend lines run the whole window.

    1. Contributors 2 +2 vs prior 12 months
    2. Maintainers 1 +1 joined · −0 left, last 12 months
    3. Bus factor 1 committers behind half the commits
    4. Community-filed 100% of issues opened, last 12 months
    5. Median time to close 11 days first maintainer reply 5.4 days
    6. Open backlog 6 of 11 issues ever filed
    1. Issues opened / month 2 · 0 communitymaintainers
    2. Open backlog 6
    3. Contributors / month 0
    4. Commits / month 0
    5. Stars 1,062
    Issue themes, last 12 months
    • other 100%

    Counts from the project's public repository and git history; no individuals are identified. Maintainers are the committers covering 80% of the last year's commits. how we measure this

    Add the badge to your README

    Maintain this project? The badge links readers straight to this page, so they can follow every release here.

    Tracked by Cyber Toolchain Choose a badge style and copy the code →
    browse all tools → score card for your toolchain →