Skip to content
Back to search

openguardrails

MarketplaceWorks with Claude CodeGitHub

The vendor-neutral protocol for AI agent safety & security — and the neutral benchmark that ranks the vendors.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 11 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/openguardrails/openguardrails with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/openguardrails/openguardrails'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/openguardrails--openguardrails/badge.svg)](/listings/openguardrails--openguardrails)

HTML

<a href="/listings/openguardrails--openguardrails"><img src="/api/listings/openguardrails--openguardrails/badge.svg" alt="findsafeskills safety rating" /></a>

Install

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 115 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. <div align="center">
  2. # OpenGuardrails
  3. **AI agent guardrails for every builder.**
  4. Let your agents do more, and walk away. Open source, and it runs on your machine.
  5. Apache-2.0 · [openguardrails.com](https://openguardrails.com)
  6. </div>
  7. ---
  8. Most of what an agent does is safe, and a rule can tell. OpenGuardrails decides those actions on
  9. your machine, with no network and no model. It sends only the ones a rule cannot decide to a model
  10. built to judge agent actions, so you stop being asked, and it tells you when something is wrong.
  11. ```text
  12. tool call -> local rules --- allow ------------------------------> runs
  13. |-- deny ---------------------------------> refused
  14. `-- cannot tell -> OpenGuardrails model
  15. |-- allow ------------> runs
  16. |-- unsafe -----------> refused, the agent is told why
  17. `-- your decision ----> the agent asks you
  18. ```
  19. Unsafe and "your decision" are different things. A key on its way to an unknown host is a security
  20. problem, and it is refused. Nobody is asked to rule on it. Buying this lipstick is safe, and it is
  21. still your call.
  22. - **Your keys stay home.** Secrets become placeholders before a request leaves your machine, side
  23. requests included, so the model provider and any relay never see them. A tool still gets the real
  24. value, and only after its call is allowed.
  25. - **Local rules.** A strict shell parser, a read-only allowlist, workspace boundaries and a small
  26. list of things no agent should ever do. A command the parser does not fully understand is never
  27. allowed by a rule. Read them: [`cli/src/gate/gate.ts`](cli/src/gate/gate.ts),
  28. [`cli/src/decide.ts`](cli/src/decide.ts).
  29. - **A refusal does not end the task.** The agent is told what was refused and what it may do
  30. instead, then carries on.
  31. - **Sandbox and firewall.** `openguardrails run` starts an agent where it cannot read your
  32. credentials and every connection is recorded. It also reports hosts the agent reached but no tool
  33. call named.
  34. - **You can see everything.** `openguardrails decisions` lists every decision, and
  35. `~/.openguardrails/model-requests.ndjson` holds exactly what was sent to the model.
  36. ## Status
  37. Early, and honest about it. Tested end to end against the real harnesses through the local gateway,
  38. with no plugin, using the scripted models in [`cli/experiments`](cli/experiments):
  39. | Agent | Rules and model enforced | Agent told why | Secrets kept from the model provider | Tool gets the real value |
  40. |---|---|---|---|---|
  41. | [Hermes](https://github.com/NousResearch/hermes-agent) 0.20 | yes | yes | yes | yes |
  42. | [OpenClaw](https://github.com/openclaw/openclaw) 2026.6 | yes | yes | yes | yes |
  43. | [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) 0.2 | yes | yes | yes | yes |
  44. Not available yet:
  45. - **The model API is not public yet.** It has been tested with the real model through all three agents, but
  46. is not deployed. Until then, what the rules cannot decide is reported as not judged and left to the
  47. agent's own permission check. Its API reference will be published with the service.
  48. - **The package is not on npm yet.** Install from a checkout, below.
  49. - Asking you on your phone. Until then, a decision that is yours comes back to you through the agent.
  50. Known limits are listed in [`cli/README.md`](cli/README.md#what-it-does-not-do-yet).
  51. ## Try it
  52. Needs Node 22.12 or newer.
  53. ```sh
  54. git clone https://github.com/openguardrails/openguardrails
  55. cd openguardrails/cli && npm install && npm run build && npm link
  56. ```
  57. Tell it about your model API in `~/.openguardrails/config.json`, mode 0600:
  58. ```json
  59. {
  60. "gateway": {
  61. "upstreams": { "main": { "baseURL": "https://openrouter.ai/api/v1", "model": "qwen/qwen3.8-27b", "apiKeyEnv": "OPENROUTER_API_KEY" } },
  62. "routes": { "auto": ["main"] }
  63. }
  64. }
  65. ```
  66. ```sh
  67. openguardrails serve # port 4180; set gateway.port in the config to change it
  68. ```
  69. Then point your agent at `http://127.0.0.1:4180/v1`, with model `auto` and any string as its API key.
  70. That is the whole integration, and it works the same for every agent.
  71. `openguardrails connect` turns on the model once it is public, and shows where to raise your quota.
  72. `openguardrails decisions` shows what happened. For an enterprise deployment, the gateway connects to
  73. OpenGuardrails AIRS instead, with no other change.
  74. ## What is in this repository
  75. | Path | What it is |
  76. |---|---|
  77. | [`cli/`](cli/) | The `openguardrails` command: rules, local service, sandbox, firewall, gateway |
  78. | [`integrations/`](integrations/) | The Higress plugin, which connects a Higress gateway to OpenGuardrails AIRS |
  79. | [`benchmarks/`](benchmarks/) | The public evaluation. We enter too, and publish the results either way |
  80. ## For companies
  81. OpenGuardrails AIRS is the enterprise product: organization-wide policy, audit, SSO, analytics and
  82. private deployment. It uses the same models. See
  83. [openguardrails.com](https://openguardrails.com).
  84. ## License
  85. Apache-2.0.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.