Deploy an AI agent three ways on Render

Prerequisites

• A Render account
• A GitHub account
• The Render CLI installed
• Node.js 22.12 or newer for the TypeScript track, or Python 3.12 and uv for the Python track
• Redis or Valkey running locally for verifying the worker ack contract (optional)
• Comfortable with Git, the terminal, and the Render Dashboard

Steps

1. 01 What you'll build Meet the code-review agent and the deploy-first plan. The agent code never changes; you deploy it three ways — in-process, queue-backed, and as Render Workflows — and watch the runtime pattern absorb more of the coordination each time. 5 min →
2. 02 Deploy the naive agent Deploy the Pattern 1 Blueprint and review a public PR on the hosted service. The review runs inside the request, which is exactly the limitation you'll feel. 5 min →
3. 03 Deploy the queue agents Deploy Pattern 2: a thin web producer, a background worker consumer, and a Key Value queue. Re-run the same public PR and watch the web service return instantly while the worker does the work. 8 min →
4. 04 Trace the ack contract Read the entry processor that decides when to acknowledge a stream message, then verify the ack boundary with a local test. 6 min →
5. 05 Scale and see the cost Scale the worker, restart the web tier mid-review, then count the coordination code you now own. 5 min →
6. 06 Deploy the workflow agents Deploy the gateway Blueprint, create the separate workflow service that Blueprints do not manage yet, then submit a public PR and read the run trace in the Render Dashboard. 9 min →
7. 07 Run workflows locally Install the Render CLI and start the local workflow runtime, then run the code-review workflow on your machine and read the same task fan-out you saw in the hosted trace. 5 min →
8. 08 Author a task Start from the shipped `your-review` sandbox, customize the `my-reviewer` agent, force a retry, then fan out a second task to see what the platform handles for you. 9 min →
9. 09 Push and choose where to run Ship the task you authored to the deployed Workflow service, verify the production trace, and confirm the same task now runs on Render. 6 min →
10. 10 Go further Challenges that add a real agentic capability on top of your review task - a judge reflection loop, an MCP-backed tool, and a human-in-the-loop gate - each one more tasks and config on the same graph. 10 min →