Agent CLI

Local AI commands for voice and text: clipboard dictation, text fixes, speech, an assistant and parallel agents in worktrees

  • 239 ★ on the repository
  • Last commit
  • MIT
CLI

Extended access

Why this level

  • Commands listen to the microphone and read the clipboard on hotkeys
  • The dev command runs several agents at once and spends tokens
  • Connecting cloud providers sends data to OpenAI or Gemini
All reasons and checks

basnijholt/agent-cli

Install

Works in

In your terminal, with SkillFoxx CLI

npx skillfoxx add cli/agent-cli

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • Commands listen to the microphone and read the clipboard on hotkeys
  • The dev command runs several agents at once and spends tokens
  • Connecting cloud providers sends data to OpenAI or Gemini

This is third-party code. Review the repository files before installing.

What it does

Agent CLI collects local AI-powered commands that run on your machine. Data from your clipboard, microphone and files stays on it, and cloud providers OpenAI and Gemini are optional. The author's idea began with dictation to the clipboard on a hotkey, to give models more context by voice. Commands: `autocorrect` fixes grammar and spelling with a local model, `transcribe` writes microphone speech to the clipboard, `speak` reads text aloud, `voice-edit` edits clipboard text by voice command, `assistant` answers a wake word, `chat` holds a conversation with tool calling, `memory` keeps long-term memory, `rag-proxy` lets you chat with documents, and `transcribe-live` transcribes in the background with voice detection. The `server` command starts local recognition and speech servers with Wyoming and OpenAI-compatible protocols. The `dev` command creates parallel git worktrees and runs AI agents in them. It installs as a skill (`agent-cli dev install-skill`) or a Claude Code plugin, and then Claude opens several agents in isolated worktrees when asked to work on several features. Short aliases `agent` and `ag` exist.

Who it is for. Developers and writers who want to dictate prompts to agents and fix text with local models without sending data to the cloud.

Good fit when

  • Typing long prompts to an agent is tiresome and you want dictation to the clipboard on a hotkey
  • Selected text should be edited by voice: make it more formal
  • You need a local recognition and speech server with an OpenAI protocol
  • Several features need parallel work, each in its own git worktree
  • You want a chat with documents through RAG without the cloud

Not a fit when

  • A cloud service with a ready subscription: the commands run on your hardware and models
  • A windowed interface with no terminal: macOS has a separate menu bar app, the rest runs from the command line
  • Local models cannot be run: quality depends on them and on your hardware

Example request

Transcribe my speech from the microphone to the clipboard with the transcribe command, then fix grammar with autocorrect

Limitations

Recognition and edit quality depend on local models and hardware: Whisper uses MLX on Apple Silicon and other mechanisms elsewhere. Background transcription needs the `agent-cli[vad]` extra. Hotkeys such as Cmd+Shift+R on macOS are set in the system. Cloud OpenAI and Gemini are optional, and then data goes to the provider. The last commit is from November 2026.

How to disable. Remove the package with uv tool uninstall agent-cli, pip uninstall agent-cli or brew uninstall --cask agent-cli. Uninstall the Claude Code skill with /plugin uninstall agent-cli-dev and delete git worktrees with git worktree remove.

FAQ

Which commands does Agent CLI have?

autocorrect, transcribe, speak, voice-edit, assistant, chat, memory, rag-proxy, transcribe-live, server and dev. Each has its own page in docs/commands/.

Is it true that data stays off the cloud?

Yes, in the default mode everything runs locally, and OpenAI and Gemini are connected only if you choose.

How do I install it?

uv tool install agent-cli (recommended), pip install agent-cli, or for the macOS menu bar app brew install --cask basnijholt/tap/agent-cli.

What does the dev command do?

It opens a separate git worktree per task and runs its own AI agent there. As a skill or Claude Code plugin it lets Claude open several agents for several features itself.

How do I install the dev skill in Claude Code?

In a project run agent-cli dev install-skill or add the marketplace with claude plugin marketplace add basnijholt/agent-cli and install claude plugin install agent-cli-dev@agent-cli.

Discussion

No questions or reviews yet.

CLI tools

Personal assistant kit for Claude Code, Codex CLI and Grok CLI: file-based memory, 23 skills and a guided first-run setup

SkillExtended access137Repository stars

A realtime voice layer on top of Claude Code, Codex and OpenCode that keeps talking and listening while background tasks run in parallel

CLIExtended access3.2KRepository stars

Voice control of Claude Code and Codex sessions on your machines: a small agent on each box links it to the pitupitu cloud.

CLIBroad access224Repository stars

Local voice layer for Claude Code and Codex on macOS: when the agent finishes or fails, it says in a sentence or two what happened

CLIExtended access217Repository stars

Open-source personal AI OS blueprint for Claude Code and Codex: context, second brain, daily briefs and drafts in your voice.

WorkflowExtended access50Repository stars

A library of 229 skills for work and life: sales, marketing, finance, engineering, design and parallel agent runs

SkillBroad access428Repository stars

Comparison with similar tools

Cards that share the main search query, and how they differ.

ToolOfficialAuthenticationRepository starsMain difference
Agent CLI (this card)no239
VoiceModeyesnot required1,454an MCP server for voice conversation with an agent; Agent CLI adds clipboard dictation, text edits and parallel worktrees
whisper.cppno55,136a speech recognition engine; Agent CLI builds commands, hotkeys and servers around it

README in short

The description presents a set of local AI command-line agents focused on privacy and offline work. The author tells how the tool grew from dictation to the clipboard, lists commands and install, and describes the dev command in the Claude Code plugin section.

README badge

Report a problem