You can connect ChatGPT to Pylon in two ways. The fastest is Pylon’s native ChatGPT app: install it from ChatGPT’s Apps tab and sign in with your Pylon account. You can also add the Pylon MCP server to ChatGPT as a custom MCP server. Either way, ChatGPT authenticates with OAuth and uses tools that read or write the same data your user can already see in Pylon.
This is a how-to for Pylon + ChatGPT. It is not a vendor roundup. For platform comparisons, see Customer support platforms that work with ChatGPT and MCP.
What you need before you start
From Pylon MCP docs:
- A Pylon workspace on a plan that includes the Pylon MCP server (listed on every plan in Features by tier).
- A Member or Admin user with the MCP Access role enabled.
- Permission for each tool operation you expect ChatGPT to run. MCP cannot return data the user cannot see, or write what the user cannot write in the Pylon UI.
- ChatGPT access that allows custom MCP servers or plugins. OpenAI documents workspace permissions and security restrictions, including Lockdown, for adding custom MCP servers (Add custom MCP server).
Pylon side: get your MCP server URL
- In Pylon, open Settings > AI Controls > MCP Server.
- Copy your workspace’s regional MCP URL from Installation.
- Use that URL in your AI tool.
Official server URLs published in docs:
- US:
https://mcp.usepylon.com - EU:
https://mcp.eu.usepylon.com
Auth is OAuth 2.0 (AuthKit). Transport is Streamable HTTP (stateless). Source: Pylon MCP.
For client-specific steps for ChatGPT, Claude, and Cursor, see connecting to the Pylon MCP server.
ChatGPT side: install the Pylon app
- In ChatGPT, open the Apps tab.
- Search for Pylon and install Pylon’s ChatGPT app.
- When prompted, sign in with your Pylon account and authorize the connection. The Pylon MCP server supports OAuth only.
- Your browser returns to ChatGPT, and Pylon’s tools are available in your conversations.
If your ChatGPT workspace restricts apps, an admin may need to allow it first.
Alternative: add Pylon as a custom MCP server
Use this path if your workspace can’t install the Pylon app or you need a custom connection.
OpenAI’s current public guide for ChatGPT on the web (Add custom MCP server; also Connect and test your plugin):
- Go to ChatGPT Plugins.
- Select the plus button, then Add custom MCP server.
- Enter a name and optional description (for example, “Pylon”).
- Under Connection, enter your Server URL (the regional URL from Pylon Installation), or use Tunnel only if you are testing a private server with Secure MCP Tunnel.
- Configure authentication. For Pylon, choose OAuth (Pylon uses OAuth 2.0).
- Review the risk warning and select that you understand and want to continue.
- Select Create as a plugin, then install the plugin before using it in a conversation.
- In a chat, type
@and select the installed plugin when you want ChatGPT to use Pylon tools.
Supported transports in ChatGPT’s custom MCP flow include SSE and streaming HTTP. Write actions require confirmation by default. OpenAI treats tools without a readOnlyHint as write actions.
Desktop note: ChatGPT desktop / Codex can also add MCP servers under Settings > MCP servers (ChatGPT Learn: MCP). Prefer the web plugin flow above for the public HTTPS Pylon server unless your team standardizes on desktop config.
What ChatGPT can do after it connects
Capabilities below come only from Pylon MCP. Summarized for buyers; the docs page is the full tool list.
Issues (read and write): search and fetch issues and messages, create and update issues, snooze, send internal notes, send customer replies, manage followers and internal threads, link product tickets (Linear, Jira, GitHub, Asana, Shortcut).
Important reply behavior: a customer reply goes out on the issue’s channel and cannot be unsent. Docs say AI tools are instructed to show the exact message and get approval before sending. Most AI tools also ask you to confirm each reply.
Accounts and contacts: search and update accounts and contacts within your permissions.
Knowledge base: search and read articles; draft, publish, and move content if you have the same edit and publish permissions as in Pylon.
Plan-gated examples:
- Account notes and some account intelligence tools require Account Intelligence.
- Tasks and projects require Account Intelligence.
- Call recordings require Account Intelligence.
- Feature Requests tools require the Feature Requests plan feature.
- Audit log search requires Admin; full history is Enterprise, otherwise last 30 days.
- AI agent backtesting tools require AI Agents and Backtesting plus MCP Access and agent permissions.
Rate limits apply per tool and organization. Throttling returns a JSON-RPC error with a retry delay.
Safe first workflows
Start with read-only work, then add writes you are willing to approve:
- “Find open issues for account Acme that mention SSO.”
- “Summarize the last 10 messages on issue 1842 for my internal note.”
- “Draft a reply, but do not send until I approve.”
- Only then: “Send the approved reply” or “Update the status after I confirm.”
Do not paste API keys into ChatGPT. Pylon MCP uses OAuth for the authenticated user.
FAQ
Does Pylon support ChatGPT over MCP?
Yes. Install Pylon’s native ChatGPT app from ChatGPT’s Apps tab, or add the Pylon MCP server as a custom MCP server. Both use OAuth and your normal Pylon permissions. Source: Pylon MCP docs.
Where do I find my Pylon MCP URL?
Open Settings > AI Controls > MCP Server in Pylon and copy the regional URL from Installation. Public base URLs are https://mcp.usepylon.com (US) and https://mcp.eu.usepylon.com (EU).
Who can connect?
Member and Admin users need the MCP Access role. Each tool still checks the user’s normal Pylon permissions.
Can ChatGPT send a customer reply from Pylon?
Yes, through send_issue_reply, subject to your permissions. Docs require showing the exact message and getting approval before send because replies cannot be unsent.
Is the Pylon MCP server on every plan?
Yes. The pricing Features by tier table lists Pylon MCP server on Professional, Advanced, and Enterprise. Some MCP tools still need add-ons or plan features such as Account Intelligence.
How is this different from your MCP platforms blog?
That post compares multiple vendors’ MCP or ChatGPT support. This post is only the setup path for Pylon and ChatGPT, with permissions and tool boundaries from Pylon docs and OpenAI’s custom MCP guide.
Ready to try it on your workspace? Schedule a demo.
