Developers
OpenAPI6 min readUpdated Aug 31, 2026

ChatGPT GPT Actions Integration Guide

Import the Orbyt GPT Actions spec to build a Custom GPT. Your GPT can read your pipeline, log activities, and help you prepare for interviews using your real data.

At a glance

What you get
All 23 Orbyt Jobs pipeline operations inside a Custom GPT: jobs, contacts, activities, calendar, pipeline summary, suggestions, and webhooks.
Orbyt plan required
Orbyt Ultra. The external API is the mcp-api feature and it is gated to Ultra, so Free and Pro keys are refused with HTTP 403.
ChatGPT plan required
ChatGPT Plus, Pro, or Team. Custom GPTs are not available on the free ChatGPT plan.
Schema URL
https://www.orbytlabs.ai/openapi-gpt.yaml, the trimmed spec. Not the full /openapi.yaml, which mixes four auth schemes and does not import cleanly.
Auth
API Key authentication with Bearer as the auth type. Orbyt Jobs tokens begin with ext_ and are created in Settings, Account, API Keys.
Write safety
Every POST, PATCH and DELETE operation is marked x-openai-isConsequential, so ChatGPT confirms with you before it changes anything.
Setup time
About ten minutes, most of it filling in the GPT's own name and instructions.

Overview

ChatGPT supports Custom GPTs with Actions, which let a GPT call external APIs on your behalf. By importing Orbyt's OpenAPI specification, you can build a GPT that reads your job pipeline, logs activities, and helps you prepare for interviews using your real data.

Prerequisites

  • Orbyt Ultra plan
  • ChatGPT Plus or Team plan (Custom GPTs require a paid ChatGPT plan)
  • An Orbyt API key (generate one in Settings > Account > API Keys)

Step 1: Generate an API Key

  1. Open Orbyt and go to Settings > Account
  2. Scroll to API Keys
  3. Click Create API Key
  4. Name it "ChatGPT" and select Read & Write permissions
  5. Copy the token (starts with ext_)

Step 2: Create a Custom GPT

  1. Go to chat.openai.com and click Explore GPTs > Create
  2. In the Configure tab, give your GPT a name like "Orbyt Job Tracker"
  3. Add instructions such as:
You are a job search assistant connected to the user's Orbyt pipeline. Use the Orbyt API to help them track jobs, prepare for interviews, and manage their career search. Always check their current pipeline before giving advice.

Step 3: Add the Orbyt Action

  1. Scroll down to Actions and click Create new action
  2. For Authentication, select API Key and choose Bearer as the auth type
  3. Paste your Orbyt API key (ext_...) as the API key value
  4. In the Schema field, click Import from URL and enter:
https://www.orbytlabs.ai/openapi-gpt.yaml
  1. ChatGPT will import all 23 Orbyt operations automatically
  2. Click Save

Use this URL, not the full /openapi.yaml. The full spec documents every Orbyt API, mixes four auth schemes, and does not import cleanly as a GPT Action. The /openapi-gpt.yaml spec is trimmed to exactly the bearer-token pipeline operations a GPT can use, with an operationId on every operation as GPT Actions require.

Step 4: Test Your GPT

Try these prompts with your new GPT:

  • "What's in my pipeline?" calls the pipeline endpoint for a summary
  • "Show me jobs I should follow up on" checks suggestions for stale applications
  • "Add a job: Product Manager at Figma, found on LinkedIn" creates a new job entry
  • "Log that I had a phone screen with Stripe today" creates an activity
  • "What interviews do I have coming up?" checks the calendar endpoint

Privacy Settings

When ChatGPT encounters an action for the first time, it will ask for your confirmation before calling the API. You can choose:

  • Always allow: ChatGPT calls Orbyt automatically (best for personal GPTs)
  • Ask each time: ChatGPT confirms before every API call (best for shared GPTs)

The OpenAPI spec marks mutation endpoints (POST, PATCH, DELETE) with x-openai-isConsequential: true, so ChatGPT will always confirm before making changes to your data.

Available Endpoints

All 23 Orbyt API operations are available through GPT Actions:

  • Jobs: CRUD operations, status updates, pipeline analytics
  • Contacts: List, create, update, delete with job linking
  • Activities: Log calls, emails, interviews, and notes
  • Calendar: View upcoming events and create new ones
  • Pipeline: Get summary with follow-up suggestions
  • Suggestions: AI agent recommendations
  • Webhooks: Manage event subscriptions

Troubleshooting

"Action not available" error:

  • Re-import the OpenAPI spec; ChatGPT may have cached an older version
  • Make sure the schema URL is exactly https://www.orbytlabs.ai/openapi-gpt.yaml

Authentication errors (401/403):

  • Verify the auth type is set to Bearer (not Basic)
  • Re-paste your API key; ChatGPT may have truncated it
  • Confirm your Orbyt plan is Ultra

GPT times out on large responses:

  • Use query parameters to limit results (e.g., ?limit=10)
  • Ask for specific data instead of "show me everything"

Tips

  • Pin your GPT to the sidebar for quick access during your job search
  • Combine Orbyt data with ChatGPT's knowledge: "Based on my pipeline and Stripe's recent earnings, help me prepare for my interview"
  • Create separate GPTs for different workflows: one for interview prep, one for pipeline management
GPT configuration (import as Custom GPT Action)Download
{
  "name": "Orbyt Job Tracker",
  "description": "Connected to your Orbyt job search pipeline. Can read your jobs, contacts, activities, calendar, and suggestions.",
  "instructions": "You are a job search assistant connected to the user's Orbyt pipeline via the Orbyt API. Use the available actions to help them track jobs, prepare for interviews, and manage their career search. Always check their current pipeline before giving advice. Be concise and actionable.",
  "actions": {
    "type": "openapi",
    "url": "https://www.orbytlabs.ai/openapi-gpt.yaml",
    "auth": { "type": "bearer" }
  }
}

Common questions

Which OpenAPI URL should I import into a Custom GPT?

Use https://www.orbytlabs.ai/openapi-gpt.yaml. That spec is trimmed to exactly the 23 bearer-token pipeline operations a GPT can call, and every operation carries the operationId that GPT Actions require. The full /openapi.yaml documents every Orbyt API across four auth schemes and will not import cleanly.

What Orbyt plan do I need for ChatGPT GPT Actions?

Orbyt Ultra. External API access is a single gated feature named mcp-api, and it is the highest tier only. A token created on Free or Pro is valid but every pipeline call it makes returns HTTP 403. You also need a paid ChatGPT plan, because Custom GPTs are not on the free tier.

Will my GPT change my job data without asking me?

No, not on the first attempt. Every mutating operation in the Orbyt GPT spec is marked x-openai-isConsequential, which tells ChatGPT to confirm with you before it runs. You can widen that to always allow in the GPT's privacy settings, which is reasonable for a private GPT and unwise for a shared one.

How many Orbyt operations does a Custom GPT get?

Twenty-three. They cover jobs with full create, read, update and delete plus status changes, contacts with job linking, activity logging, calendar reads and writes, a pipeline summary with follow-up suggestions, agent recommendations, and webhook subscription management.

Why does my GPT get a 401 or 403 from Orbyt?

A 401 means the token is wrong, revoked, or was truncated on paste, and re-pasting it usually fixes it. A 403 means the token is valid but the Orbyt account behind it is not on Ultra. Check that the auth type is set to Bearer and not Basic, which is the other common cause of a 401.

Is it safe to share a Custom GPT that is connected to my Orbyt account?

Treat the GPT as a credential. The token you paste has read and write access to your entire pipeline, so anyone who can run the GPT can act on your data through it. For anything shared, create a separate read-only token and set the GPT's privacy setting to ask each time.

My GPT times out on large responses. What should I do?

Ask for less. Most Orbyt list endpoints accept a limit query parameter, so a prompt like show my ten most recent applications returns a response the GPT can actually read. Requests for everything are the usual cause, because the model has to consume the whole payload before it answers.

ChatGPT says the action is not available. How do I fix it?

Re-import the schema. ChatGPT caches the spec it fetched when you created the action, so a stale copy survives changes on our side. Confirm the URL is exactly https://www.orbytlabs.ai/openapi-gpt.yaml, delete the action, and add it again rather than editing it in place.

Other integration guides

Ready to build?

Ship it.