Pylon MCP Server

Official Pylon MCP server. Read and write issues, accounts, and contacts in your Pylon customer support workspace via OAuth.

Collaboration by Pylon OAuth2 active

Connect (Remote Server) Docs

Overview

Pylon's official MCP server is a provider-hosted remote endpoint at https://mcp.usepylon.com that lets AI tools like Claude, ChatGPT, and Cursor read and write data in your Pylon customer support workspace. It exposes 11 tools spanning issues (search, retrieve, message history, create, update), accounts (search, retrieve, update), and contacts/users (get contact, get user, get authenticated user). Access is scoped to the same visibility you have in the Pylon dashboard, and write actions respect your existing UI permissions.

The server uses streamable HTTP transport (stateless) and authenticates via OAuth 2.0 through AuthKit. Connecting from a Member or Admin seat with the "MCP Access" permission enabled is required, and the feature is toggled by an admin in Settings → AI Controls → MCP Server. Rate limits are applied per tool, with 429 responses returned when limits are hit.

Because Pylon hosts the server itself, there is no repository to clone or package to install. Users simply point their MCP client at the URL and complete the OAuth flow in the browser. This makes it suitable for ad-hoc customer analysis, debugging issues with code generation, or generating product specs from issue collections.

Tools

Tool	Description
`search_issues`	Search Pylon issues across your accessible workspace.
`get_issue`	Retrieve a single issue by ID with full details.
`get_issue_messages`	Fetch the message history for a specific issue.
`create_issue`	Create a new issue in Pylon.
`update_issue`	Update fields on an existing issue, including status, assignee, and custom fields.
`search_accounts`	Search for accounts in your Pylon workspace.
`get_account`	Retrieve a single account's details by ID.
`update_account`	Update fields on an account.
`get_contact`	Retrieve contact information by ID.
`get_user`	Retrieve team member (user) information by ID.
`get_me`	Return details for the authenticated user.

Setup Guide

Prerequisites

A Pylon workspace with a Member or Admin seat
"MCP Access" permission enabled on your role
An admin must enable the MCP Server in Settings → AI Controls → MCP Server

Connect from Claude Code

claude mcp add pylon --transport http https://mcp.usepylon.com/

Then run /mcp inside Claude Code and select the pylon server to start the OAuth flow.

Connect from Claude Desktop / Claude Web

Open Settings → Connectors
Add a custom connector with URL https://mcp.usepylon.com/
Leave OAuth credentials blank
Click Connect and complete the browser OAuth flow

Connect from Cursor

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "pylon": {
      "url": "https://mcp.usepylon.com/"
    }
  }
}

Then authenticate via Cursor Settings → Tools & MCP.

Connect from ChatGPT

Search for Pylon in the Apps tab and install the native application.

Authentication

OAuth 2.0 via AuthKit. The browser opens automatically when the client connects, you sign in with your Pylon account, authorize the connection, and the session is established. Sessions expire periodically and require re-running the connect flow. API keys are not supported.

Use Cases

Triage and update support issues from chat: search open issues, read message history, then update status, assignee, or custom fields without leaving the AI client
Generate product specs from a collection of related Pylon issues by searching and reading their message threads
Run ad-hoc customer analysis by pulling account details and the issues attached to specific accounts
Create new issues directly from an AI workflow (for example, file a bug or follow-up from a meeting transcript)
Look up contact and team member context (get_contact, get_user, get_me) before drafting customer replies

Example Prompts

"Search Pylon for open issues mentioning 'SSO' and summarize the top 5."
"Get the full message history for issue ISS-1423 and draft a reply based on the latest customer message."
"Find the Acme Corp account in Pylon and list all issues attached to it."
"Create a new Pylon issue titled 'Webhook retries failing' assigned to me with the details from this conversation."
"Update issue ISS-987 to status 'In Progress' and assign it to the on-call engineer."

Pros

Official, provider-hosted server with no install or self-hosting required
OAuth 2.0 authentication scopes access to exactly what the user can see in the Pylon UI, including write permissions
Covers the core support workflow: issues (including message history and creation), accounts, and contacts/users
Works with major MCP clients out of the box: Claude Desktop, Claude Code, Cursor, and ChatGPT

Limitations

Limited surface area: 11 tools covering issues, accounts, and contacts/users only. No tools for tags, knowledge base articles, teams, or analytics
Per-tool rate limits with 429 responses; not suited for bulk migrations or large analytical exports
Requires admin enablement plus a Member/Admin seat with MCP Access permission; not available to all roles or API-key-only environments

Alternatives

Intercom MCP server for conversational customer support workflows
Zendesk MCP integrations for ticket and account management
Pylon REST API directly, via a generic HTTP MCP wrapper, when you need data outside the 11 exposed tools