Sentrook: FIDU Agentic Security Tool
We’re building collective security for open agents.
Sentrook is the security layer that lets open-source and community-built agents compete with closed systems. We’re building security infrastructure with absolute loyalty to your interests, not to a profit centre.
With Sentrook, your open agent gets
- Protection against attacks designed by the adversary community (not just internal teams)
- Real-time updates as new threats emerge
Right now, the agent landscape is split.
Closed
Closed systems (Gemini, Anthropic, OpenAI) own the agents and the security.
Open
Open agents like Open Claw have innovation, speed and freedom, but little collective-intelligence led security.
Sentrook provides what the Open ecosystem needs to compete.
Built for Performance and Flexibility
Fast
Striving for extremely low latency that you won’t even notice
Adaptable
Built with OpenClaw as the first target, but designed to work with any agent, with many more integrations planned
Smart gating
Protects your agent’s tool calls only when necessary based on our ever-growing ruleset. No preconfigured manual blacklists, restrictive allow sets, or complex configuration
Safe data handling
Plan data is meticulously scrubbed and anonymised before it ever leaves your machine
The cost of being unprotected?
Compromised tools, stolen memory, data exfiltration: the kinds of attacks that look like bugs until they’re weaponised.
We all want security, and only a tool that harnesses the social signal can deliver that.
Open Source, Free and Membership-based access
We’ve thought hard about how a truly open source philosophy should apply to an AI-agent security tool.
Sentrook is open source and can be run by anyone. A free FIDU membership will give access to our hosted instance and the community-driven attack ruleset.
The data on threat descriptions and rules generated from them are held by FIDU, a not-for-profit, and are not made open source in order not to facilitate circumventing attacks
Free tier
A free tier of protection is available which provides:
- Rule-based assessment with frequent updates of the rules for new attack vectors, in exchange for
- Securely shared data of suspected attacks
What’s next
Initially, Sentrook will use rules and community-driven examples to gate and protect your agent’s tool calls, driven by the community’s data. But we already have big plans going forward, like offering FIDU Members hot-path LLM assessment to help catch even novel attacks, where membership dues cover the model costs.
Your data, your choice
You also get the option not to share back your security information to strengthen the commons of the ecosystem.
(We would, of course, discourage you from doing this unless you have very particular security needs.)
Run it yourself
Sentrook is on GitHub
The source and full install instructions live in the repository.