Skip to main content

Deploy as ChatGPT

Quick Steps (TL;DR)

1. Deploy your agent
2. Copy its OpenAPI URL
3. Paste it into ChatGPT → Add Actions → Import

When to Use ChatGPT Deployment in SmythOS?​

Deploying as ChatGPT allows you to connect your SmythOS agent with a custom GPT inside ChatGPT. By importing the OpenAPI schema generated by SmythOS, you can expose your agent's capabilities as actions that ChatGPT can call on demand.

This creates powerful use cases:

  • Extend ChatGPT with your domain-specific logic
  • Let users interact with your workflows via natural language
  • Control agent behavior via secure, versioned APIs

Prerequisites​

Before you start, you’ll need:

  • A deployed SmythOS agent (Test or Production)
  • Access to OpenAI GPT Builder (Pro or higher)

Step 1: Copy Your Agent's OpenAPI URL​

Your Agent's OpenAPI schema is the bridge to ChatGPT. It's a JSON file that tells GPT Builder what your Agent can do.

  1. In your SmythOS dashboard, navigate to your Agent's Settings → Deployments.
  2. Under the Environment URLs section, copy the subdomain for your chosen environment.
https://[your-agent].agent.stage.smyth.ai
https://[your-agent].agent.pstage.smyth.ai
  1. Create the full OpenAPI URL by appending /api-docs/openapi-llm.json to the subdomain you copied.
https://[your-agent].agent.pstage.smyth.ai/api-docs/openapi-llm.json
Test vs. Production Environment

Use the Test URL for development and iteration, as it always reflects your latest unpublished draft. Use the Production URL for stable, public-facing GPTs.

Step 2: Build Your Custom GPT​

1. Open GPT Builder​

Click here to open GPT Builder

2. Set Basic Configuration​

Click the Configure tab.

3. Add Name, Instructions, and Welcome Message​

Fill in:

  • Name – the public-facing name of your GPT
  • Instructions – paste your SmythOS agent behavior or prompt config
  • Welcome message – what users see on launch
INFO
For consistency, paste the same instructions or system prompt used in your agent.

4. Add Your Agent as an Action​

  • Scroll to the Actions section
  • Click Create new action
  • Paste your OpenAPI URL (Test or Production):

Test Agent:​

https://your-agent.agent.stage.smyth.ai/api-docs/openapi-llm.json

Production Agent:​

https://your-agent.agent.pstage.smyth.ai/api-docs/openapi-llm.json
  • Click Import to load the schema.

The GPT Builder will fetch your agent’s API definition and list available actions.

URL Format Note

Make sure the URL resolves directly to a .json file. If redirected, use the final URL.

6. Save and Test​

  • Click Create in the top right to save your GPT
  • Interact with it like any other GPT, but now powered by your SmythOS agent
  • Optionally, share the GPT with others via a public link
Success

You’ve connected your agent to ChatGPT! Your workflows now power a custom GPT.

Keeping Your GPT Synced​

Live Sync with Test Mode

If you’re using a stage subdomain, your GPT reflects the latest edits. In debug mode, ChatGPT waits for manual approval in Studio.

  • Re-deploy your agent after edits
  • Re-check your subdomain to ensure it’s unchanged
  • Re-import the OpenAPI URL in GPT Builder to sync new logic
Reimport your OpenAPI URL after every significant agent update or version bump.

What’s Next?​

Great! Your SmythOS agent is now live and available inside ChatGPT as a custom GPT.

  • Build different GPTs using the same backend logic
  • Route traffic by versioning your agent and updating GPT URLs
  • Monitor usage from Settings → Deployments → Analytics in SmythOS
🔧 Troubleshooting Common Issues
IssueFix
401 UnauthorizedDouble-check auth token format and value.
Import failsVisit the URL in a browser. If it redirects, use the final .json link.
GPT doesn't see new endpointsRe-import the OpenAPI file after making changes.
('openapi',): Input should be '3.1.0'Update the OpenAPI version from 3.0.1 to 3.1.0 in the schema.