Unified command center for AI coding tools
Manage Claude, Gemini, Qwen, and 22 more AI assistants from one terminal interface.

Safe Testing Recommended: Terminal Jarvis is a harness for AI coding tools that can modify files and execute commands. For the safest experience, test in a remote development environment such as GitHub Codespaces, Coder, DevPod, or Google Colab.
A data-driven harness switcher for AI coding agents. Maps 25 coding-agent CLIs through a shared 9-capability contract -- one interface to download, run, update, and inspect any agent tool.
# Cargo
cargo install terminal-jarvis
# npm
npm install -g terminal-jarvis
# Homebrew
brew install BA-CalderonMorales/homebrew-terminal-jarvis/terminal-jarvis
Cargo builds the Rust CLI from the crates.io source package. The npm package is
a Node launcher that downloads the matching Terminal Jarvis GitHub Release
asset, verifies its .sha256 file, caches it, and then executes it. Homebrew is
the binary installer path and installs the platform release archive from the tap.
Supported prebuilt assets are linux-x64-gnu, macos-x64, and macos-arm64.
Native Windows npm installs are not published yet; use WSL on Windows.
# List every coding agent
terminal-jarvis list
# Inspect a harness
terminal-jarvis show opencode
# Preview a capability command
terminal-jarvis plan codex headless
# Select and verify the active harness
terminal-jarvis use opencode
terminal-jarvis current
terminal-jarvis check
For development builds, replace terminal-jarvis with cargo run --.
harnesses/<agent>/
├── index.toml # name, display, binary, env requirements
├── download/index.toml # install without sudo
├── update/index.toml # upgrade without interactive auth
├── headless/index.toml # non-interactive command mode
├── version/index.toml # print installed agent version
├── stats/index.toml # local agent statistics
├── models/index.toml # list available models
├── security/index.toml # sandbox and approval settings
├── ui/index.toml # interactive terminal UI
└── yolo/index.toml # bypass safeguards (dangerous)
Auth stays with each harness -- terminal-jarvis never retains credentials.
| Command | Purpose |
|---|---|
list |
Show all coding agents |
show <harness> |
Inspect a harness's capabilities |
use <harness> / current |
Select / show active harness |
plan [harness] <capability> |
Preview the shell command |
run [harness] [capability] [args...] |
Execute a capability |
check |
Report binary + env readiness |
security [status\|audit\|harness] |
Security posture |
version [--verbose] / --version / -v / --info |
Version info |
config show |
Active config state |
auth help <harness> |
Credential setup guidance |
[harness] [args...] |
Pass-through to harness binary |
Legacy aliases from 0.0.x: see docs/migration.md.
| Document | What |
|---|---|
| Capability contract | Full breakdown of the 9 capabilities |
| Supported agents | All 25 coding agents |
| Maintainer guide | Development, verification, release |
| Architecture | Module boundaries and contracts |
| Release plan | Checklist and auth boundaries |
| Migration | Breaking changes from 0.0.x |
| Testing | Test gates and integration hardening |
| Distribution hardening | Package integrity and provenance |
$ claude mcp add terminal-jarvis \
-- python -m otcore.mcp_server <graph>