Caliber: Open-Source Proxy for Enforcing LLM Agent Rules

GlobalDevelopersCreated Apr 27, 2026, 6:40 AM

Cross-posting here because this problem affects everyone building with AI agents. Prompt-based guardrails fail. The model follows your system prompt in a demo, then ignores rules when context gets big or the agent chains multiple steps. We built Caliber - an open-source proxy that reads your rules from plain markdown and enforces them at the API layer, not in the prompt. Every call. Provider-agnostic. Just hit 700 GitHub stars ⭐ and nearly 100 forks - the reception from devs building with AI has been amazing. Repo: https://github.com/caliber-ai-org/ai-setup Would love: \- Feedback on the approach \- Feature requests from people building AI agents \- Anyone who wants to contribute to the project Building this open-source for the community.

Caliber: Open-Source Proxy for Enforcing LLM Agent Rules

Caliber is an open-source proxy designed to enforce rules for large language model (LLM) agents, ensuring that their responses align with predefined guidelines and ethical standards. This tool is particularly valuable in scenarios where maintaining control over generated content is crucial.

Use Cases

Content Moderation : Implementing Caliber helps in filtering out inappropriate or sensitive content, making it ideal for platforms that host user-generated content.
Regulatory Compliance : Companies in regulated sectors can use Caliber to ensure that their LLM agents comply with local and international laws and regulations.
Brand Safety : Businesses can enforce brand-specific guidelines to ensure that all generated content aligns with their brand voice and values.

Pros of Using Caliber

Customizable Rules : Caliber allows for the easy enforcement of specific rules that can be customized according to the user’s needs.
Open-Source : Being open-source, Caliber offers transparency and flexibility, allowing developers to modify and extend its functionality.
Scalability : The proxy system can handle large volumes of data, making it suitable for high-traffic platforms and applications.

FAQ

What is Caliber used for?

Caliber is used to enforce rules on the outputs of LLM agents, ensuring that the generated content adheres to specific guidelines and standards.

How do I get started with Caliber?

You can start by accessing the Caliber repository on GitHub. Follow the installation and setup instructions provided in the documentation.

Can Caliber be customized to fit specific needs?

Yes, Caliber is highly customizable. You can define your own rules and guidelines to tailor the enforcement mechanism to your specific requirements.

Is Caliber suitable for large-scale applications?

Yes, Caliber is designed to be scalable and can handle large volumes of data, making it suitable for high-traffic platforms and applications.

Is there a community or support available for Caliber?

As an open-source project, Caliber benefits from community support. You can find additional resources, forums, and potential contributors on the project's GitHub page and other community platforms.

By integrating Caliber into your systems, you can maintain greater control over the outputs of your LLM agents, ensuring consistency, compliance, and brand safety.

Caliber: Open-Source Proxy for Enforcing LLM Agent Rules