Preview. These maps are our reading of public sources and haven't been checked with the projects' maintainers yet.
numbat (Perplexity)
Endpoint detection and response for AI agents: collects activity from agent hooks, OTLP logs and on-disk session files, runs CEL detection rules locally, can optionally block via pre-action hooks, and reconstructs sessions forensically.
v0.3.0 released 2026-10-19; repo pushed 2026-10-25; created 2026-07-24. (checked 2026-10-25)
In Agent guardrails
Where it stands apart
- How is it mainly sustained? big company side project (0% of 20 others)
- Which shared standard or format does it mainly build on? coding agents' hook interfaces (Claude Code, Codex, Cursor hooks) (0% of 21 others)
- Where does it mainly step in? outside the action path (inventory, records, or checks afterwards) (14% of 21 others)
- If the guard itself breaks, is missing, or cannot decide, what happens by default? fails open: the action runs unguarded (12% of 17 others)
- What is the main thing it protects? the truth of the record (what the agent did or claimed) (10% of 21 others)
What it shares with most of the community
- Is it actively maintained? active (81% of 21 others)
- How is the main thing licensed? permissive open source (MIT/Apache/BSD) (67% of 21 others)
- Can it run entirely on your own device or infrastructure, without the maker's servers? yes (76% of 21 others)
- Do you need an account to use it? no account (67% of 21 others)
- Who is it for? developers (90% of 21 others)
- What mainly decides whether an action is allowed? rules, policy, or boundaries written in advance (57% of 21 others)
- Which agents does it work with? any agent or process (not limited to a named list) (71% of 21 others)
- In the default setup, where does the agent's work actually execute? directly on the host machine, as usual (52% of 21 others)
- Does it control the agent's outbound network access? no (56% of 16 others)
- What is it mainly for? adding checkpoints, review, or evidence (52% of 21 others)
- Is enforcement kept out of the agent's reach? yes, enforced in code or infrastructure the agent cannot change (95% of 21 others)
Nearest neighbors
- Vultrino 71% of 17 shared answers match
- cMCP (AgenTrust) 67% of 15 shared answers match
- OpenAPPA 67% of 15 shared answers match
Every answer
| Question | Answer | Source |
|---|---|---|
| Is it actively maintained? | active | source 2026-10-25 |
| How is the main thing licensed? | permissive open source (MIT/Apache/BSD) Apache-2.0 | source 2026-10-25 |
| How is it mainly sustained? | stands apart big company side project Published by Perplexity. | source 2026-10-25 |
| Can it run entirely on your own device or infrastructure, without the maker's servers? | yes | source 2026-10-25 |
| Do you need an account to use it? | no account Single local binary; install from release or go build, rules run locally; no sign-in in README. | source 2026-10-26 |
| Who is it for? | developers Security and platform teams. | source 2026-10-25 |
| Which shared standard or format does it mainly build on? | stands apart coding agents' hook interfaces (Claude Code, Codex, Cursor hooks) Collects from each agent's hooks (plus OTLP); blocks via pre-action hooks. | source 2026-10-26 |
| Where does it mainly step in? | stands apart outside the action path (inventory, records, or checks afterwards) Monitor-only by default; blocking is opt-in through pre-action hooks. | source 2026-10-25 |
| What mainly decides whether an action is allowed? | rules, policy, or boundaries written in advance CEL and YAML rules; all shipped rules are monitor-only. | source 2026-10-25 |
| If the guard itself breaks, is missing, or cannot decide, what happens by default? | stands apart fails open: the action runs unguarded Errors, panics and oversized input 'suppress the numbat deny'. | source 2026-10-25 |
| Which agents does it work with? | any agent or process (not limited to a named list) Many desktop, CLI, IDE and gateway agents per its coverage matrix. | source 2026-10-25 |
| In the default setup, where does the agent's work actually execute? | directly on the host machine, as usual | source 2026-10-25 |
| What is the main thing it protects? | stands apart the truth of the record (what the agent did or claimed) Primarily visibility and forensics. | source 2026-10-25 |
| Does it keep a record of agent actions? | yes, ordinary logs or a timeline Versioned NDJSON records; case bundles carry SHA-256 manifests. | source 2026-10-25 |
| Does it control the agent's outbound network access? | no | source 2026-10-25 |
| What is it mainly for? | adding checkpoints, review, or evidence | source 2026-10-25 |
| Is enforcement kept out of the agent's reach? | yes, enforced in code or infrastructure the agent cannot change | source 2026-10-25 |
Sources
This page was derived from:
- GitHub · github.com/perplexityai/numbat · read 2026-10-25
- GitHub · github.com/perplexityai/numbat/blob/main/docs/enforcement.md · read 2026-10-25
RSS for this project · Its own release feed
Are you a maintainer? Claiming and correcting projects is coming soon.