Saif Ali

Posted on Mar 29

How to Deploy OpenClaw in NEXUS AI

#ai #openclaw #devops #tutorial

OpenClaw Personal AI Assistant Deployment

OpenClaw
is a lightweight AI coding gateway — Claude Code-compatible — exposes a secure token-authenticated endpoint. NEXUS AI makes deploying it entirely conversational: no YAML files, no Dockerfiles, no console wrestling. Just ask and it's live.

This guide walks you through deploying OpenClaw on NEXUS AI step by step — covering provider selection, environment setup, connecting your CLI, and optional advanced configuration. By the end you'll have a live gateway you can connect Claude Code or any compatible AI coding tool to.

What is OpenClaw?

OpenClaw (alpine/openclaw:latest) is a minimal, production-ready AI gateway service designed to work with Claude Code and compatible clients. It exposes a single authenticated endpoint on port 18789, identified by a gateway token that's either auto-generated or supplied by you at deploy time.

Once deployed, any Claude Code-compatible tool can point at your OpenClaw instance and authenticate with the token — giving you a self-hosted, cloud-native AI coding backend.

Prerequisites

A NEXUS AI account with at least one project configured The NEXUS AI MCP connector enabled in Claude — or API access to api.zollo.live/mcp A supported cloud provider account (for GCP, AWS, or Azure deployments) The OpenClaw CLI installed locally to connect after deployment Tip: You can also deploy OpenClaw entirely from within a Claude chat session if you have the NEXUS AI MCP connector enabled. No terminal required for the deploy itself.

Deploying OpenClaw — step by step

01 Choose your provider & environment NEXUS AI supports four deployment providers out of the box. For local development and testing, Docker is the fastest path. For production workloads, GCP Cloud Run offers the best cold-start performance and auto-scaling.

🐳 Docker Best for local dev & CI

☁️ GCP Cloud Run Serverless, auto-scales

⚡ AWS ECS Fargate Full AWS ecosystem

🔷 Azure Container Apps Enterprise & compliance

Environments map to: DEVELOPMENT for local testing, STAGING for pre-production validation, and PRODUCTION for live traffic.

02 Trigger the deployment Via the NEXUS AI MCP in Claude, call nexusai_deploy_openclaw with your provider and environment. A minimal invocation looks like this:

nexusai_deploy_openclaw(
  provider:    "gcp_cloud_run",
  environment: "PRODUCTION",
  name:        "openclaw-gateway"  // optional
)

If you don't supply a gatewayToken, one is securely auto-generated for you. You can also pass in Claude API credentials (claudeApiKey, claudeWebCookie) if your gateway requires them.

03 Save your deployment credentials After a successful deploy call, NEXUS AI returns:

Field Description Example id Deployment UUID 40693d12-… gatewayToken Auth token for CLI 6fb631a9d6… status Current state queued → running url Public endpoint Available once live Important: Copy your gatewayToken immediately — it's only returned once. Treat it like a password and store it in a secrets manager or your .env.

04 Monitor deployment status Deployments move from queued → provisioning → running. Check status anytime:

nexusai_deploy_status(
  deploymentId: "40693d12-0839-4821-aa3c-3621f12c74f4"
)

Once status shows running, the url field will be populated with your public endpoint.


- 05 Connect via the OpenClaw CLI
Point your local tooling at the gateway by setting two environment variables:

## Set your gateway token

shell
export OPENCLAW_GATEWAY_TOKEN="6fb631a9d6c66c45cb3f5890886b14dc"

Point CLI at your deployment URL
export OPENCLAW_GATEWAY_URL="https://your-deployment-url


## Verify the connection
openclaw status
You should see a confirmation that the gateway is reachable and authenticated. Claude Code and any compatible AI coding client will now route through your self-hosted OpenClaw instance.

* Tip: For persistent config, add these exports to your .zshrc / .bashrc, or use a .env file with direnv.
Advanced configuration
Custom gateway token
If you want a predictable, rotatable token instead of an auto-generated one, pass it explicitly at deploy time:

NEXUS AI MCP
nexusai_deploy_openclaw(
provider: "gcp_cloud_run",
environment: "PRODUCTION",
gatewayToken: "my-secure-token-from-vault"
)
Passing Claude credentials
If your OpenClaw instance needs to authenticate directly with Anthropic, you can pass your Claude session credentials at deploy time via claudeApiKey, claudeWebCookie, or claudeWebSessionKey. These are stored securely in NEXUS AI's secrets system and injected into the container at runtime.

Scaling your deployment

Once your OpenClaw gateway is live, you can scale it up instantly via NEXUS AI without redeploying:

nexusai_deploy_scale(
  deploymentId: "40693d12-…",
  replicas:     3
)```


## Custom domain
Attach a custom domain to your OpenClaw gateway for a clean, branded endpoint:
NEXUS AI MCP nexusai_domains_add( deploymentId: "40693d12-…", domain: "openclaw.yourdomain.com" )

Managing your OpenClaw deployment

NEXUS AI gives you full lifecycle control over every deployment. Here's a quick reference:

Action MCP Tool When to use Check status nexusai_deploy_status After deploying, or to verify health View logs nexusai_deploy_logs Debug connection or auth issues Stop gateway nexusai_deploy_stop Pause to save compute costs Restart nexusai_deploy_start After a stop, or after config changes Rollback nexusai_deploy_rollback Revert to a previous revision Delete nexusai_deploy_delete Permanently tear down

Troubleshooting

Gateway stuck in "queued" Check that your cloud provider credentials are properly configured in NEXUS AI. For GCP, ensure your service account has Cloud Run admin permissions. Use nexusai_deploy_logs to see what's happening inside the container.

CLI returns 401 Unauthorized
Double-check your OPENCLAW_GATEWAY_TOKEN environment variable. Tokens are case-sensitive. If you've lost the token, delete the deployment and redeploy with a custom gatewayToken you control.
Connection refused on port 18789
Verify the deployment status is running (not queued or stopped). Also confirm your OPENCLAW_GATEWAY_URL includes the correct port and protocol.

Wrapping up

OpenClaw on NEXUS AI gives you a fully managed, cloud-native AI coding gateway without any of the infrastructure overhead. You get auto-generated tokens, multi-cloud flexibility, instant scaling, and full lifecycle management — all through natural language or a single API call.

From here, you can explore connecting multiple Claude Code clients to the same gateway, attaching a custom domain, or setting up a staging → production promotion workflow entirely within NEXUS AI.

Try it now: Head to https://nexusai.run/. and enable the NEXUS AI MCP connector in Claude to deploy your first OpenClaw gateway in under 2 minutes.

DEV Community