We are excited to announce the release of the Stigg Agent Governance SDK, an open-source solution built on top of Stigg's highly scalable infrastructure, that empowers you to seamlessly meter, enforce, and govern every LLM call your product and its agents make!

🆕 What’s new?

Managing autonomous AI agents requires granular control over limits, costs, and access. The new Agent Governance SDK solves this by wrapping the LLM or gateway client you already use. By simply calling stigg.wrap(client) once, every subsequent model call goes through three critical steps:

1. Meter usage: Token counts (and costs, where supported) are automatically read from each response and reported to Stigg straight from the provider, making usage metering completely hands-off. The SDK includes native support for 18 LLM providers and gateways, including: OpenRouter, LiteLLM, Cloudflare, Databricks, Amazon Bedrock, OpenAI, Anthropic, Google Gemini, Groq, Mistral, Deepseek, Perplexity, Ollama, TypeSafe (Jev) and more. Supporting additional vendors is as easy as initializing the SDK with a recipe that defines the mapping logic.
2. Enforce access: Before a call even leaves your environment, the SDK checks that the agent has enough credits or metered capacity to perform its task. Agents that have hit their limit are blocked before they reach the provider, preventing unexpected costs.
3. Govern usage: Give each agent its own identity and budget under your teams and organizations. You’ll know exactly which agent spent what, and you can stop a single runaway agent without disrupting the rest of your system.

💡 Why it matters

• Native support for common LLM providers and gateways: Building custom metering logic takes time away from your core product. Our LLM and LLM gateway wrapping capabilities intercept calls natively, extracting usage and cost metrics automatically so you don't have to. 
• Enterprise-grade scale: The SDK is powered by Stigg's enterprise-grade infrastructure. Enjoy peace of mind with high-throughput metering capable of ingesting up to 1M+ events per second, and low-latency access enforcement that returns decisions in under 10ms.
• The power of Open Source: We've made the Agent SDK fully open source. This gives your engineering team complete transparency into how your LLM requests are handled in the request path, the flexibility to inspect and contribute to the code, and the freedom to easily customize the SDK for proprietary endpoints or internal architectures.

⚒️ Get started

1. Install the Agent SDK: npm install @stigg/agent-sdk-js
2. Read the Agent SDK documentation
3. Explore the open source library

📦 Availability

The Agent Governance SDK is available in public beta for all customers.

🍿 See it in action

Watch how simple it is to add metering, access enforcement, and governance to your AI agents: