Sleeping
2
Kedi Playground
🐈
Run Kedi code with customizable AI models in the browser
Agentic Programming Language for typed LLM Orchestration.
Kedi is a typed orchestration language for building LLM programs whose prompts, dataflow, structured outputs, tools, tests, and agent configuration live in one readable source format. Use it through .kedi programs or embed the same runtime semantics through its Python API.
It combines runtime substitution and typed output capture with reusable procedures, deterministic Python, model and agent configuration, tools, MCP, skills, approvals, subagents, tests, evaluations, and prompt optimization. Model calls run through framework adapters such as Pydantic AI, DSPy, or LangChain, or harness adapters such as Codex, Claude, and ACP.