Claude Code
Anthropic
Local desktop workspace
Agent Launcher
Six coding agents. One local entry point.
One-click installOne-click configureOne-click launch
Detect local CLIs automatically: install only what is missing, link what already exists, then choose an API profile to start a session or resume one from local history.
Filter by associated directory~/Projects/agent-launcher
Resume a previous sessionRead from local session history
Improve installer detectionToday 10:42 · ~/Projects/agent-launcher
Review native configurationYesterday 18:06 · ~/Projects/agent-launcher
Existing CLIs stay intactNative config syncLocal session historyOpen source under MIT
One entry point, six agents
Keep the CLIs you know. Lose the repeated setup.
Agent Launcher still runs each project's own command-line tool. Existing installations keep their paths, config homes, and session history.
Codex CLI
OpenAI
OpenCode
SST
Pi
earendil
Gemini CLI
Hermes Agent
Nous Research
Real application screens
From linking to sessions, the workflow stays in one window.
01 / 04
Projects and session history
Browse each CLI's local history and associated directories, then start in the embedded terminal or resume an existing session.
Filter by associated directory~/Projects/agent-launcher
Resume a previous sessionRead from local session history
Improve installer detectionToday 10:42 · ~/Projects/agent-launcher
Review native configurationYesterday 18:06 · ~/Projects/agent-launcher
Get started
One-click install, configure, and launch.
01
One-click install
Scans your system for CLIs. If one is missing, the app uses that CLI's official installation path; existing commands are linked without reinstalling or upgrading.
02
One-click configure
Pick an official account or compatible API profile. The app writes the environment variables and native config files the CLI reads, then verifies connectivity.
03
One-click launch
Start a new session in the embedded terminal or chat view, or resume an existing one from local history.
Built for daily work
It does not replace your CLIs. It organizes them.
Existing installs stay intact
Detection runs before every install attempt. If a command exists, Agent Launcher links it without reinstalling, repairing, or upgrading.
Native config sync
File-backed config for Claude Code, Codex, OpenCode, Pi, and Hermes; environment-only setup for Gemini.
History across formats
Read each CLI's own JSONL, JSON, or SQLite history and resume with its native command arguments.
Explicit control for risky modes
Auto-approval is opt-in per supported CLI. Pi has no redundant toggle because it does not prompt for approval.
Local-first, not offline theater
Local data stays local. Model requests go where you choose.
Agent Launcher reads sessions, configuration, and usage on your machine and has no separate analytics service. Running an agent sends requests to the official provider or relay in the active profile; version checks may contact GitHub, npm, or PyPI.
About API keys
By design, keys are stored as plaintext in ~/.agent-launcher/config.json and, when required, native CLI config files. The interface masks them, but your local user account can still read those files.
Put six agents in one workspace.
Get the macOS, Windows, or Linux package from GitHub Releases.
FAQ
Clear boundaries make better tools.
Will one-click install overwrite my existing CLIs?+
No. Agent Launcher links existing commands first. One-click install appears only when a CLI cannot be found; it never upgrades or modifies existing installations.
Where is data stored?+
Application state lives under ~/.agent-launcher/. System-linked CLIs keep using their own configuration and history directories.
Which packages are supported?+
macOS has DMG and ZIP builds for Intel and Apple silicon, Windows uses NSIS, and Linux uses AppImage. Available files are listed on GitHub Releases.
Is it completely offline?+
No. Configuration, history, and usage are handled locally, but agent requests go to your selected model service and version checks can reach public software sources.