Quickstart

Sign-in happens in the browser. You never paste a Cloud key into the terminal to authenticate.

1. Install

curl -fsSL https://raw.githubusercontent.com/RunanywhereAI/wally/main/install.sh | sh

If you would rather not pipe a script to a shell, take the signed release for your platform from the releases page instead. macOS on Apple Silicon and Windows x64 have published artifacts.

wally account login

The console opens with a short-lived request. Sign in with Google or GitHub, check the hostname in the browser, and approve only a terminal you just started yourself. The credential lands in the terminal after you approve. If the browser cannot open, rerun with --no-browser and open the printed URL yourself. Never put an access or refresh token in a URL, shell history, issue, or log.

3. Run against the cloud

Hosted and local are separate choices, so ask for the hosted one:

wally opencode --cloud --model <model-id>

The model catalogue on console.runanywhere.ai lists the ids your signed-in account can pass here. A Cloud key is entitled separately, so the /v1/models response in step 4 is scoped to that key and need not match the account catalogue — read the list for whichever credential you are calling with. Arguments for the harness itself go after --, as in wally opencode --cloud --model <model-id> -- --help.

4. Call the API directly

Cloud usage bills against prepaid credits. The terms and the refund policy are the commercial contract. Create a Cloud key on the console’s Cloud keys page. The plaintext shows once, so put it straight into your secret manager, and never reuse an on-device runa_prod_ SDK key here. Ask the API which models the key can call:

export RUNA_CLOUD_KEY='sk-runa-…'

curl https://inference.runanywhere.ai/v1/models \
  -H "Authorization: Bearer $RUNA_CLOUD_KEY"

That response is the authoritative list. Take an id from it for <model-id> below and in step 3:

curl -N https://inference.runanywhere.ai/v1/chat/completions \
  -H "Authorization: Bearer $RUNA_CLOUD_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"model":"<model-id>","stream":true,"messages":[{"role":"user","content":"Say hello in one sentence."}]}'

The response is OpenAI-compatible SSE and ends with data: [DONE]. See the API reference for request fields and error behaviour.

Troubleshooting

Status	What it means
`401`	The Cloud key is missing, expired, revoked, or is a device SDK key
`403`	The key is valid but that model or route is not entitled, or the account is out of credit (`insufficient_credit`)
`429`	Capacity or rate limit. Wait the number of seconds in `Retry-After` when the response carries one, and back off exponentially with jitter when it does not

No console URL on wally account login usually means the installed release is stale. Update from the releases page rather than copying an old localhost command.

Get started

API reference

Reference

1. Install

3. Run against the cloud

4. Call the API directly

Troubleshooting

​1. Install

​2. Sign in

​3. Run against the cloud

​4. Call the API directly

​Troubleshooting

1. Install

2. Sign in

3. Run against the cloud

4. Call the API directly

Troubleshooting