Product
Compare and Run Any Agent Framework from One Interface
any-agent is the universal interface for evaluating and building across today’s leading agent frameworks: standardized, traceable, and integrated with the Mozilla.ai Agent Platform.
Why any-agent?
Building and debugging agents is hard, each framework has its own API, with different input and output formats.
any-agent normalizes tracing and evaluation across frameworks using GenAI OpenTelemetry standards. This means you can debug, compare, and iterate faster, without rewriting your stack.
Supported Frameworks
One Interface, Many Frameworks
Build agents once, and switch frameworks with a configuration change update, no code rewrites required.
Standardized OTEL GenAI Tracing
No matter which framework you use, the agent trace and input/output logging structure is the same. This simplifies the process of monitoring agent behavior without needing to rely upon framework specific designs.
Trace-First Evaluation
LLM-as-a-judge and Agent-as-a-judge evaluation techniques allow for easy analysis of the entire agent trace, to understand and evaluate each step the agent took during execution.
Support for MCP and A2A
Integrates with Mozilla.ai’s mcpd and supports multi-agent communication via MCP and Agent2Agent protocols, out of the box.
See it in Action
Location-Based Assistant Demo
An example built with any-agent that integrates retrieval, LLM calls, and tool-based planning to find and recommend the best surf spots.
Hosted by
Get Started with any-agent
Choose a supported framework, configure your project, and run one of our ready-to-use examples.