Connect your tools

Connect OpenClaw

Use your Straitly key in the OpenClaw app and in the terminal (openclaw). One paste sets up both, and the models you pick show up in OpenClaw’s model menu.

App + CLI setup
Have ready
OpenClaw installed and curl available. The screenshots follow the Mac app; the setup command uses the shell rather than Python.
What changes
A separate ~/.openclaw/straitly.json file and an entry in OpenClaw's secret store. Existing settings are kept.

Keep copied keys and setup commands private.

Before you start

The command in step 3 creates a key for you, or uses one you already have. You need the OpenClaw app or openclawinstalled. You can skip the app’s setup screens: the command does that part for you.

Setup

1. Pick your models

These are the models you will see in OpenClaw. The first one is picked when it opens.

Claude Opus 5.5GPT-6 AstraClaude Fable 5.1

Price per million tokens, input / output. Only models that can run OpenClaw are listed.

2. Open Terminal

Press ⌘ Space, type Terminal and press Enter.

Spotlight search with Terminal selected

3. Paste the command

Copy it, paste it into Terminal and press Enter. You should see Done - OpenClaw now uses Straitly.

It puts Straitly in its own file, ~/.openclaw/straitly.json, and keeps the rest of your OpenClaw settings. Your key goes into OpenClaw’s secret store, not into a file. It works right away. If you only just installed the OpenClaw app, it also finishes the app’s setup, which takes about a minute.

Terminal after pasting the command, showing Done - OpenClaw now uses Straitly

4. Open OpenClaw

Open the OpenClaw app. It goes straight to a chat, with your first model picked. In the terminal, run the command the paste printed at the end, usually openclaw chat.

5. Pick a model

Click the model name under the message box. Your models are in the Straitly group. In the terminal, type /model instead.

6. Send a message

Say anything. If OpenClaw answers, you are done.

Seeing the thinking

The command turns on OpenClaw’s reasoning setting, so when a model thinks, its thinking arrives as a separate message. In the app, also turn on View → Reasoning.

With Claude, do not type /reasoning onOpenClaw sends that text to Claude, and Anthropic can refuse chats that ask Claude to show or talk about its thinking. The reply fails with Agent run failed, and later messages in that chat can fail too. Reasoning is already on. If a chat keeps failing, start a new one with /new.

Change your models later

Pick again in step 1 and paste the new command. It replaces the list and works right away. Every id is on the models page.

Verify your connection

Open the app or run the chat command printed by setup. Choose a Straitly model, send a short message, and check Straitly Activity.

Open request activity to check the model, status, and charge for your test request.

Setup checks

This is a fresh app installation

Let setup finish. Completing the app's initial setup can take about a minute; wait for the Done message before opening a chat.

A Claude conversation keeps failing

See the reasoning warning above. Do not send /reasoning on to Claude; use a new chat with /new if that conversation keeps failing.

If the request is rejected, check your balance and key limits in Errors & limits.

Undo

This takes Straitly out of OpenClaw, puts back the model you used before and leaves the rest of your settings as they are:

bash
curl -fsSL https://straitly.ai/setup/openclaw | STRAITLY_UNDO=1 sh