weio-cli 0.5.0

Weio — an agentic coding assistant that routes inference through your Weio account.

weio-cli

An agentic coding assistant that runs on your machine and routes inference through your Weio account.

Install

pip install weio-cli        # or: pipx install weio-cli

Requires Python 3.9+.

Authenticate

weio login            # opens your browser, sign in (incl. Google), key is created automatically

weio login starts a one-time local handshake, opens weio.ai in your browser to sign in, mints an API key for this device, and saves it to ~/.weio/config.json. No copy-paste.

Prefer to paste a key yourself? Generate one in Settings → API & CLI on weio.ai and:

weio login --no-browser     # paste your weio_sk_… key
# or, per session:
export WEIO_API_KEY="weio_sk_…"

Check what you're using:

weio usage            # tier, tokens used today, remaining, reset time

Use

Interactive agent (default)

Run weio with no arguments to start an autonomous coding agent in your terminal — like Claude Code / Codex. It reads files, searches, edits, and runs commands to complete your task, showing each step:

cd my-project
weio

weio › add a /health route to app.py and run the tests

● Let me look at the app first.
  📖 read_file app.py
● Adding the route.
  ✎ edit_file app.py
  │ + @app.get("/health")
  │ + def health(): return {"ok": True}
  apply edit_file to app.py? [y/N] y
● Verifying.
  ❯ run_command pytest -q
  │ exit=0  …  5 passed
✔ Added /health and confirmed tests pass.

The agent uses tools: read_file, list_dir, search, edit_file, write_file, run_command. File paths are confined to the project directory.

Slash commands: /model <id>, /approval <mode>, /steps <n>, /settings, /undo, /clear, /cwd, /help, /exit.

Approval modes

Control how much the agent does without asking (/approval or weio config approval):

Mode	Edits	Shell commands
suggest (default)	ask each time	ask each time
auto-edit	auto-apply	ask each time
full-auto	auto-apply	auto-run

Settings

weio config                      # show all settings
weio config approval auto-edit   # set a value
weio config model pro            # auto | low | mid | pro
weio config max_steps 40

Settings live in ~/.weio/settings.json. The agent uses the most capable model your plan allows.

One-shot & other commands

# Run a coding task non-interactively (auto-applies edits; add --auto to also run commands):
weio "add error handling to the fetch() in api.py"
weio code "refactor db.py to async" --auto

# Quick question (no file edits):
weio ask "what does a 502 from nginx usually mean?"

weio chat        # plain interactive chat
weio usage       # tokens used / limits
weio ping        # connectivity + key check

Edits are shown as a diff and require confirmation before anything is written (use -y/--yes to apply automatically). New files are created as needed.

Updating

weio-cli checks PyPI once a day (fail-silent) and prints a one-line notice when a newer version is available. To upgrade:

weio update           # upgrades in place via pip
# or
pip install -U weio-cli

Disable the check with WEIO_NO_UPDATE_CHECK=1.

Configuration

Setting	Flag	Env	Config file
API key	--key	WEIO_API_KEY	~/.weio/config.json
API base	--base	WEIO_BASE	~/.weio/config.json
Model	--model	—	—

Self-hosted / LAN gateway? Point at it with --base http://HOST:8901/v1.

Output is billed against your Weio account usage.