Build Reliable AI Agents with Simple YAML
Liman is a declarative, language-agnostic framework for building AI agents using YAML manifests.
Built for AgentOps at Scale
From OpenAPI integration to distributed execution, Liman provides everything needed to build, deploy, and operate production AI agents across multiple languages and environments.
OpenAPI → Tools Generation
Automatically generate LLM tools from OpenAPI specifications. Transform any API into agent tools without writing MCP servers.
Service Account Authorization
Built-in authorization with service accounts and role assumption. Secure access control for distributed agent execution.
Dynamic Prompt Localization
Multi-language support with automatic system prompt generation. Increase function calling accuracy across languages.
OTel & FinOps Out-of-Box
Built-in OpenTelemetry integration with cost tracking. Monitor performance, token usage, and financial metrics automatically.
Distributed Edges
Connect nodes via MCP, A2A, HTTP, WebSocket, or shared memory. Build distributed agents across AWS Lambda and processes.
Atomic State Management
External state with pre_hook, invoke, post_hook phases. Build complex distributed agents with flexible state handling.
Plugin Ecosystem
Extensible plugin system with built-in and custom plugins. Auto-context stitching, evaluation agents, and anomaly detection.
Kustomize Overlays
Layer configurations using Kustomize-like overlays. Perfect for multi-environment deployments and language variants.
Stay Updated
Get the latest updates on Liman development, new features, and best practices delivered to your inbox.
No spam, unsubscribe at any time. We respect your privacy.
Join the Community
Participate in specification development and share experiences with other developers