Open Think Deploy →
§13 · Guided deploy · terminal walkthrough

Deploy Helm with your hands —
without the install.

Each step looks like a wrangler command you'd run locally. Click [Run] and we execute the equivalent against the Cloudflare API on your behalf. Same result, no npm install -g wrangler required, full audit trail. For the form-style version, see /deploy/cloud.

helm@open-think — guided deploy
$ wrangler login

# wrangler login normally opens your browser to OAuth. We don't have a backend
# PTY, so paste a scoped API token instead — create one ↗


        
$ wrangler whoami

# Shows the accounts your token can deploy to. Pick the one you want.


        
$ cat > wrangler.toml <<EOF

# Configure your deployment. Deploys at <name>.<your-account-subdomain>.workers.dev.

# Optional model keys

        
$ wrangler deploy

# Provisions D1 + Access (if enabled), composes wrangler.toml,
# sets every secret you supplied, uploads the Worker bundle.


          
        
§13.1 · How this works under the hood

Real wrangler is local. We're remote.

Each click runs the equivalent Cloudflare API call from this Worker (open-think.app), authenticated with your token. Nothing executes on your machine. Same result as running wrangler locally: a script in your account, secrets in CF's secret store, your D1 bound.

We don't spin up a per-session container or a real PTY — projects like vercel-labs/wterm give you a beautiful in-browser terminal UI but still need a backend with Node + wrangler installed. For the deploy use-case, going directly through the CF API is simpler, fully audited, and skips 50 MB of WASM payload.

If you want the real wrangler experience, the form-style page at /deploy/cloud still emits a paste-ready wrangler.toml + a single wrangler deploy for you to run locally.