OpenClaw + EvoLink API Setup Guide in 2026: One Gateway for Chat, Image, and Video Models
npm install -g openclaw@latest path, followed by openclaw onboard --install-daemon. EvoLink's current integration docs then give you a practical route for plugging OpenClaw into a single API key workflow.TL;DR
- OpenClaw's current docs recommend either the website installer or
npm install -g openclaw@latest. - The current onboarding flow is
openclaw onboard --install-daemon. - EvoLink's current docs position it as a single gateway for multiple model families, which is useful when OpenClaw should not be tied to only one provider.
- The safest rollout is: install OpenClaw, run onboarding, connect a channel, then switch the provider configuration to EvoLink.
What the current docs say
| Topic | Current documented guidance |
|---|---|
| OpenClaw install | Website installer or npm install -g openclaw@latest |
| Onboarding | openclaw onboard --install-daemon |
| Verification | openclaw doctor, openclaw status, and openclaw dashboard are part of the current install docs |
| Channel setup | OpenClaw docs describe logging into channels after onboarding |
| EvoLink integration | EvoLink docs describe OpenClaw as a gateway workflow that can use EvoLink as the model provider |
Why teams pair OpenClaw with EvoLink
OpenClaw solves the channel and agent side of the workflow. EvoLink solves the provider-side sprawl.
That combination is useful when you want:
- one OpenClaw deployment serving multiple model families
- fewer provider-specific code changes
- the option to route among chat, image, and video models through one API key
- a simpler setup for Telegram or Feishu style integrations
Recommended setup flow
1. Install OpenClaw
If you want the quickest path, use the official website installer documented by OpenClaw. If you already manage Node and global packages yourself, the docs also support:
npm install -g openclaw@latest
openclaw onboard --install-daemon2. Run onboarding first
Do not edit everything by hand before the base install works. The official docs are built around the onboarding flow, and that is still the fastest way to create a valid baseline config.
3. Connect a channel
OpenClaw's current getting-started flow expects you to connect a channel after install. Depending on your workflow, that may be Telegram, Feishu, or another supported channel.
For a personal or lightweight team setup, Telegram is usually the fastest place to validate the whole stack.
4. Point your model provider to EvoLink
After the base OpenClaw environment works, switch the provider-side configuration to EvoLink and use your EvoLink API key as the upstream credential.
At a high level, the provider split looks like this:
| Layer | Role |
|---|---|
| OpenClaw | Channel connection, session flow, tool orchestration |
| EvoLink | Upstream model gateway and API surface |
| Model routes | Claude, GPT, Gemini, image, and video families behind the gateway |
5. Verify before you scale
The OpenClaw docs already point to a good verification path:
- run
openclaw doctor - run
openclaw status - open
openclaw dashboard - send a real test message from your chosen channel
Do this before adding multiple bots, users, or workflows.
A simple production checklist
| Check | Why it matters |
|---|---|
| OpenClaw installs cleanly | Avoids debugging provider config on top of a broken local install |
| Onboarding completes | Confirms the base service layout is valid |
| Channel login succeeds | Proves your messaging layer is actually connected |
| EvoLink API key is loaded correctly | Prevents silent upstream auth failures |
| Test request returns a real model response | Validates end-to-end path before rollout |
FAQ
What is the current recommended way to install OpenClaw?
npm install -g openclaw@latest, followed by onboarding.Do I still need Node.js?
Yes. OpenClaw's current install docs say Node is required, and the npm-based install path assumes a working Node environment.
What command starts the onboarding flow?
openclaw onboard --install-daemon.When should I switch OpenClaw to EvoLink?
After the initial OpenClaw install and onboarding are working. It is easier to debug provider configuration once the base environment is already healthy.
Why not configure everything manually from the start?
Because onboarding is the shortest route to a known-good baseline. Manual edits are easier after that.
Is this setup only for Claude?
No. The useful reason to place EvoLink behind OpenClaw is to keep room for multiple model families through one gateway path.
Start with the Guided Route
If you want a faster OpenClaw-to-EvoLink path, EvoLink's current OpenClaw materials provide a guided setup route for Telegram and Feishu workflows.
Open the EvoLink OpenClaw Setup Flow
