ainativelang

Why we included this project

Most agent teams end up re-prompting an LLM to decide routing, state, and next steps on every run of a recurring job. AINL is built around a different idea: it is a compact, graph-based language that compiles a workflow once into a deterministic intermediate representation, and the same source can then target runtimes like LangGraph, Temporal, or FastAPI without being rewritten. Multi-step agents with memory, tool use, and validation gates execute repeatedly without burning tokens on orchestration each time. The project ships as an MCP server, so Claude Code, Cursor, Cline, and other MCP-capable agents can author and run these workflows directly, and compile-time validation plus tamper-evident execution traces cover teams with compliance requirements. If your scheduled monitors, digests, or other pipelines currently lean on long prompt loops, this is a practical way to make them structured and repeatable.