AgentUse CLI
Agents for recurring work in Markdown files: schedules, approval of important actions, run history in a dashboard and a Mac app
- 208 ★ on the repository
- Last commit
Extended access
Why this level
- Agents run tools and shell commands on a schedule without you present, and approvals set the boundaries
- Model provider keys are stored in the app and CLI settings
- The local `agentuse serve` HTTP interface must not be left open to the outside
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/agentuseDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Agents run tools and shell commands on a schedule without you present, and approvals set the boundaries
- Model provider keys are stored in the app and CLI settings
- The local `agentuse serve` HTTP interface must not be left open to the outside
This is third-party code. Review the repository files before installing.
What it does
AgentUse keeps an agent in an `.agentuse` file with Markdown instructions, tools, a model and a schedule. The agent does recurring work in the background: research, reports, preparing materials, work reminders. Before an important action it stops and waits for a human decision, and dangerous shell commands can pass only through an approval enforced by the runtime. It runs from a Mac app with a bundled runtime, from the CLI with a web dashboard on Windows, Linux or macOS (Node.js 22 or newer is required), or on a server, in Docker or CI. The dashboard shows output, artifacts, tool calls, usage and approval history, and past sessions turn into proposed agent edits. An agent can be created through Codex, Claude Code or the setup wizard.
Who it is for. Managers and analysts in small teams who need an agent for recurring scheduled work where a human keeps the final say.
Good fit when
- Every morning a repository digest is needed and an agent must prepare it on a schedule
- An agent drafts an email or a post and a human confirms sending in the dashboard
- Run results must be kept: output, artifacts, tokens and approval decisions
- The work has to live on a server while a laptop is off
Not a fit when
- An interactive helper in an editor is needed: AgentUse is built for background work
- Instant answers are needed with no history and no schedule
- The license matters: the repository page lists it as non-standard
Example request
Make an agent that collects a digest of my repositories every morning and asks for approval before posting to a channel.Limitations
The CLI needs Node.js 22 or newer, the Mac app includes its own runtime and the choice of model provider is yours. Schedules work while the app or server runs on a machine that is awake. The license in GitHub metadata is marked NOASSERTION, so read the terms in the `LICENSE` file. The latest commit in the repository is from October 2026. Commands such as `agentuse serve` open a local HTTP interface, so do not expose it outside without protection.
How to disable. Run `npm uninstall -g agentuse` or remove the Mac app, then stop `agentuse serve` and delete the schedules from the `.agentuse` files. The agent files and session history sit in your project folders.
FAQ
How do I install AgentUse?
Run `npx -y agentuse@latest setup` for the wizard or `npm install -g agentuse`, then `agentuse provider login`. On Apple silicon a Mac app with a bundled runtime is available.
Where are agents stored?
In `.agentuse` files: Markdown instructions, tools, a model and a schedule sit together, so one file runs on a laptop, a server, in CI and in a container.
How do approvals work?
The agent stops before an important action and you accept, reject or request changes in the dashboard. Shell commands have a mode with approval enforced by the runtime.
Can an agent be called over HTTP?
Yes. After `agentuse serve` the local `/api/run` address accepts requests, and a schedule is set in the agent file frontmatter.
Which models work?
The provider is chosen during setup and the model is set with `--model`, for example `openai:gpt-5.6`. The `agentuse models` command lists the available ones.
Discussion
No questions or reviews yet.
Related
CLI toolsA Rust CLI for macOS: an agent reads any app's interface through the accessibility tree and clicks, types and scrolls by element refs, with no screenshots
A self-hosted control plane for running several agents: task dispatch, run review, spend tracking, connected via MCP, CLI or REST
CLI dispatcher for agents: tasks run in the background and the agent sleeps on a question until you reply from a messenger.
A Swift MCP server that lets an agent drive any macOS app: click, type and read the interface tree
Turns one run of a task into a program an agent calls as a single tool, checking the record before and after the save
groundcrew hands Linear and Jira tasks to local Claude Code, Codex and Cursor agents, each in its own worktree and sandbox
README in short
The README states the goal, explains three ways to run (Mac app, CLI with web dashboard, server), shows an example agent and the list of triggers: command, schedule, HTTP. Models, approvals and documentation follow.