InitRunner ✓

You've built agents before. You know how it goes — install a framework, write a model client, define tools in code, bolt on memory, figure out RAG, add logging, then do it all again for the next agent.

InitRunner compresses that into a YAML file.

Install the package, define your agent's model, system prompt, tools, and guardrails in a role.yaml, and run it from your terminal. You configure agents in YAML instead of writing the wiring code yourself.

One role file, four ways to run it: as a one-shot command, an interactive REPL, an autonomous agent that plans and executes multi-step tasks, or an OpenAI-compatible API server.

RAG, long-term memory, and token budget management are declared in the same YAML — not stitched together in application code. Nine providers work out of the box (OpenAI, Anthropic, Google, Groq, Mistral, Cohere, Bedrock, xAI, Ollama). Switching is one line.

Agents ship with tools for the things you'd otherwise build yourself: Git operations, filesystem access, HTTP requests, web scraping, shell commands, SQL queries, Python execution, email, web search, Slack messaging, scheduling, and MCP server
integration. Custom tools drop in as a single file. Every interaction is recorded in an append-only audit log. A web dashboard and terminal UI let you monitor runs without digging through stdout.

When one agent isn't enough, a compose.yaml wires multiple agents into a single workflow — with routing, shared context, and structured output between them.