InitRunner

InitRunner is an open-source CLI and dashboard that defines AI agents in a single YAML file and runs them with one command. It adds chat, RAG, memory, 28 tool types, audit logs, guardrails and support for 12+ model providers.

InitRunner is an open-source tool for defining and running AI agents from the terminal. Instead of writing framework code, a user describes an agent in a single YAML file: a name, a prompt, a model, the tools it may use and any guardrails. One command then starts it. The project is released under MIT OR Apache-2.0, installs with a shell script, uv or pip, and needs Python 3.11 or newer. The target reader is a developer or operations engineer who wants agents that behave like configuration. Files can be checked into git, validated before use and swapped between providers by changing one line. Supported providers include Anthropic, OpenAI, Google Gemini, Groq, Mistral, Cohere, AWS Bedrock, xAI, Ollama for local models, OpenRouter, DeepSeek, Azure OpenAI and any OpenAI-compatible endpoint. A setup command stores the provider key, and running with no file opens a plain chat session. In practice the feature set goes well beyond chat. Agents get 28 built-in tool types (filesystem, HTTP, MCP, shell, git and search among them) and can load custom Python functions. Pointing a flag at a folder builds a retrieval index over markdown, PDFs or CSVs without a separate vector database. Memory comes in semantic, episodic and procedural forms, and reusable skills live in SKILL.md files. Reasoning strategies such as plan and execute run under enforced limits on tokens, tool calls, cost and time. Triggers (cron, file watchers, webhooks, Telegram, Discord and Slack) let agents run unattended in daemon mode, and an agent can be exposed as an OpenAI-compatible API. Every input, tool call and output lands in an append-only SQLite audit log, with OpenTelemetry available for tracing. Among alternatives, InitRunner sits between code-first frameworks such as PydanticAI and LangChain and hosted no-code agent builders. It can import agents written in those two frameworks, offers a built-in web dashboard with an agent launchpad and flow builder, and shares agent definitions through a public hub called InitHub. Teams that prefer visual tooling or fully managed hosting will find it more hands-on, while those who want auditable, version-controlled agents on their own infrastructure get a compact option.