Automating with the Cloud API

Use the Cloud API to trigger agents from build systems, cron jobs, or other services—no manual intervention required.

Example: GitHub Actions workflow


name: testbase
 
on:
  push:
    branches: [main]
 
jobs:
  run-cloud-agent:
    runs-on: ubuntu-latest
    env:
      TESTBASE_API_KEY: ${{ secrets.TESTBASE_API_KEY }}
      TESTBASE_API_URL: https://testbase-cloud-api-310716408703.us-central1.run.app
    steps:
      - uses: actions/checkout@v4
 
      - name: Install dependencies
        run: npm install computer-agents @testbase/cloud
 
      - name: Execute agent
        run: node ./scripts/run-cloud-agent.mjs

run-cloud-agent.mjs:


import { TestbaseCloud } from '@testbase/cloud';
import { run } from 'computer-agents';
 
const cloud = new TestbaseCloud({
  apiKey: process.env.TESTBASE_API_KEY,
  apiUrl: process.env.TESTBASE_API_URL
});
 
const agent = await cloud.createCloudAgent({
  name: 'CI Worker',
  agentType: 'worker',
  workspace: './',
  persistContainer: false
});
 
await cloud.uploadFile(agent.containerId, '/workspace/.env', process.env.APP_ENV || '');
 
const result = await run(agent, 'Run tests and summarize failures');
console.log(result.finalOutput);
 
await cloud.deleteContainer(agent.containerId);

Example: raw HTTP integration


API="https://testbase-cloud-api-310716408703.us-central1.run.app"
KEY="tb_prod_..."
 
# Create container
container=$(curl -s -X POST "$API/api/v1/containers" \
  -H "Authorization: Bearer $KEY" \
  -H "Content-Type: application/json" \
  -d '{
        "agentType": "worker",
        "workspace": "/workspace",
        "model": "gpt-4o-mini",
        "reasoningEffort": "low"
      }')
 
id=$(echo "$container" | jq -r '.id')
url=$(echo "$container" | jq -r '.cloudRunServiceUrl')
 
# Upload bootstrap file
curl -X PUT "$API/api/v1/containers/$id/files" \
  -H "Authorization: Bearer $KEY" \
  -H "Content-Type: application/json" \
  -d '{ "path": "/workspace/README.md", "content": "# Hello" }'
 
# Kick off task via container endpoint
curl -X POST "$url/execute" \
  -H "Content-Type: application/json" \
  -d '{ "task": "Generate release notes" }'
 
# Fetch artefacts
curl "$API/api/v1/containers/$id/files?path=/workspace" \
  -H "Authorization: Bearer $KEY"
 
# Clean up
curl -X DELETE "$API/api/v1/containers/$id" \
  -H "Authorization: Bearer $KEY"

Automation tips

Retry idempotent operations: container creation, file uploads, and MCP updates can be retried safely if you receive 5xx responses.
Tag runs: store CI build IDs in your own database keyed by containerId to map outputs back to commits.
Parallelize carefully: each container has isolated storage, but remember that API keys govern quotas. Spread large batches across multiple keys/environments.
Alert on health: poll GET /api/v1/containers/:id/health in long automations and surface alerts when healthy flips to false.

Automation turns TestBase into a programmable platform—integrate it wherever you run scripted workflows.