Clawsec.bot

Clawsec is an open-source security plugin for OpenClaw that intercepts and blocks dangerous tool calls from autonomous AI agents in under 5ms.

What is Clawsec?

Clawsec is an open-source plugin for OpenClaw that adds real-time security guardrails. It intercepts every tool invocation (shell commands, HTTP requests, file writes), scores risk against built-in rulesets, and enforces configurable actions (block, confirm, allow). Developed by Clawdsec, it is MIT licensed and available on GitHub.

Key Features

• Sub-millisecond latency — Risk scoring completes in under 5ms.
• Six built-in rule categories — Destructive Commands, Secret Exposure, Data Exfiltration, Unauthorized Purchases, Privilege Escalation, and Prompt Injection.
• Three enforcement actions — block, confirm (human approval), or allow + log.
• YAML configuration — Extend or override rules via clawsec.yaml with built-in rule sets like builtin/aws-security.
• One-command install — openclaw plugins install clawsec; zero additional config required.
• Open source — MIT licensed, view source on GitHub.

Who is it for?

• Developers building autonomous agents — Add protection against prompt injection and secret leakage without writing security code.
• DevOps engineers — Enforce audit trails for all agent tool calls with block/confirm policies.
• OpenClaw users in production — Install the plugin to get guardrails out of the box, customizable for their environment.

Use cases

• Prevent file deletion — When an agent interprets "clean up old logs" as rm -rf /var/log/*, Clawsec blocks the command and notifies the user.
• Stop secret leakage — Catches API keys (e.g., AKIA*) before they are written to logs or sent externally.
• Require approval for financial transactions — Configure confirm to wait for user approval on payment API calls.

How does it work?

Clawsec intercepts each tool call at the middleware layer, runs it through a rules engine that matches patterns (e.g., destructive-commands/rm-recursive), and assigns a risk score. Based on the configured action, it blocks, asks for human confirmation, or allows the call with a full audit trail — all in under 5ms.

Pricing

Clawsec is free and open source under the MIT license. There is no paid tier or usage limit.

FAQ

Is Clawsec free?

Yes, it's MIT-licensed open source. Install with one command, no cost.

How do I install it?

Run openclaw plugins install clawsec. The plugin auto-generates a default clawsec.yaml.

What does it protect against?

Out of the box: destructive commands, secret exposure, data exfiltration, unauthorized purchases, privilege escalation, and prompt injection. All rules are configurable.