About Agentuse
AgentUse is an open source agent runtime for recurring work. Agents are defined as plain Markdown files, with model settings and capabilities in YAML frontmatter and the instructions written in natural language, and AgentUse supplies the operating layer around them: the agent loop, tools, durable sessions, schedules, approval gates, HTTP endpoints and an operations dashboard.
It suits teams that want to hand off repeat jobs such as weekly reports, invoice collection or change monitoring while keeping consequential actions under review. Agents can run from the command line, on a schedule, from an HTTP trigger or event, in CI or inside Docker, and setting approval to true pauses a run before actions like sending, publishing or deleting until a reviewer approves, rejects or comments, after which the same session resumes. Models can come from Anthropic, OpenAI, OpenRouter, OpenCode Go, Amazon Bedrock or a compatible local endpoint and can be overridden at runtime. Agents can use allowlisted filesystem and shell access, built in tools, reusable skills, subagents, stores and Model Context Protocol servers. Because definitions live in the repository, they can be diffed, reviewed and versioned like code, and the vendor publishes workflow guides with prompts for building agents using Codex or Claude Code. It ships as a Mac app for Apple silicon, plus a CLI and web dashboard for Windows, Linux and macOS that require Node.js 22 or later.
The project is released under the Apache License 2.0 and no hosted control plane is required. No paid plans or prices are published on the site or in the documentation.