Ageniti

Ageniti is a developer framework designed for building applications whose capabilities can be exposed as structured actions across multiple environments. Its core model centers on defining actions as the primary unit of agentic capability, allowing developers to create a single action contract and then generate adapters for various surfaces, including CLI, HTTP, MCP (Machine Control Protocol), OpenAI and AI SDK tools, React hooks, and typed clients. This approach enables the same application logic to be accessed and operated by agents, operators, or automation systems without duplicating definitions for each interface.

The framework provides a runtime that handles critical production concerns such as validation, permissions, retries with per-attempt cancellation, idempotency, concurrency limits, secret redaction, and detailed error codes. Ageniti supports streaming events as a primitive, so logs, progress updates, and artifacts can be emitted live and subscribed to by any consumer, such as user interfaces, agent loops, or log shippers. Integration is facilitated through practical starter patterns and recipes, making it easier to wire application features to different surfaces like CLI, MCP servers, React apps, and automation scripts.

Ageniti is unopinionated about higher-level orchestration, leaving decisions about prompt orchestration, tool selection, and workflow composition to the developer. It focuses on providing the action contract and runtime, while orchestration, memory management, and workflow composition remain outside its scope. Developers can bring their own schema validation libraries, including Zod, Valibot, ArkType, or any library implementing Standard Schema v1, ensuring compatibility with existing typed validation.

The framework is delivered as the @ageniti/core package via npm, and is built on technologies such as React, TypeScript, Node.js, and Next.js. Documentation, templates, and a GitHub repository are available to support integration and onboarding. Ageniti is intended for developers seeking to expose application capabilities to agents and automation tools through explicit, reviewable contracts and reusable product capabilities.