Choice-first Stack - Build AI with the tools of your choice

Build, evaluate, and guard your AI agents and apps with composable, interoperable tools — all chosen by you.
We offer a unified open-source stack that simplifies building and testing modern AI agents and apps.Streamline your entire development workflow – no more juggling fragmented tools – and focus on creating intelligent, reliable, and production-ready apps within a seamless, interoperable ecosystem.
Build lightweight, observable AI agents with a minimal Python framework.
A unified interface for working with different LLM providers. Flexible, not complex.
A unified interface to multiple guardrail models. No extra engineering required.
Simplify MCP servers management across all environments. Close the dev-to-prod gap.
Whether you’re contributing to open-source or deploying at scale, these tools are your gateway to flexible, trustworthy AI.
Choose your agent using tinyagent.
Route LLM calls via any-llm.
Define the tool servers declaratively with mcpd.
Guard your agent’s outputs and reasoning by calling any-guardrail.
Iterate or swap components as needed without rewriting orchestration logic.
An AI-readable version of this website is available in llms.txtPortions of this content are ©1998–2023 by individual mozilla.org contributors.MZL.AI, PBC, 1875 Mission Street, San Francisco CA 94103, USA