Common Room MCP Server

Overview

Common Room's MCP server connects your AI assistant directly to your go-to-market data — product signals, community activity, intent data, contact enrichment, and more. Using the Model Context Protocol (MCP), any compatible AI client can query your Common Room workspace in natural language to research accounts, prep for calls, build prospect lists, and draft personalized outreach.

The server is hosted at https://mcp.commonroom.io/mcp

Prerequisites

Before connecting any AI client to Common Room's MCP server, you'll need a Common Room instance and an account with access to the instance.

Connecting Your AI Client

Claude

1. Open Claude's settings and navigate to the Connectors → Manage section
2. Install the Common Room connector
3. Complete the OAuth authorization flow to connect your Common Room account
4. Common Room tools will be available in your Claude conversations immediately

ChatGPT

Common Room is available as an official ChatGPT App, providing a one-click connection experience.

Install via the ChatGPT App Store:

1. Visit the Common Room ChatGPT App and click Add to ChatGPT
2. You'll be redirected to authorize Common Room with your account credentials
3. Once connected, open a new chat and enable the Common Room connector via the + menu in the message bar → More → select Common Room

Cursor

1. Open Cursor and go to Settings (⌘, on Mac / Ctrl+, on Windows)
2. Navigate to Tools & Integrations → MCP
3. Click Add MCP Server
4. Select HTTP as the transport type and enter the server URL:
   <https://mcp.commonroom.io/mcp>
5. Follow the OAuth prompt to authenticate with your Common Room account
6. The Common Room tools will now be available in Cursor's Agent and Composer panels

You can also configure the server directly in your ~/.cursor/mcp.json file:

{
  "mcpServers": {
    "common-room": {
      "type": "http",
      "url": "<https://mcp.commonroom.io/mcp>"
    }
  }
}

GitHub Copilot (VS Code)

To add Common Room to your user profile (available across all workspaces):

1. Open the Command Palette (⇧⌘P / Ctrl+Shift+P) and run MCP: Open User Configuration
2. Add the servers configuration below.
3. Save the file — VS Code will prompt you to trust the server
4. Common Room tools will now appear in GitHub Copilot Chat when using Agent mode

{
  "servers": {
    "common-room": {
      "type": "http",
      "url": "<https://mcp.commonroom.io/mcp>"
    }
  }
}

To add it to a specific workspace only, create or edit .vscode/mcp.json in your project root with the same configuration above.

Other MCP-Compatible Clients

Any client that supports the MCP specification over HTTP can connect to Common Room's MCP server. Use the following configuration details:

Clients confirmed to support remote HTTP MCP servers include Windsurf, Zed, Amazon Q Developer, Microsoft Copilot Studio, Vercel AI, and any client built on the official MCP SDK. Refer to your client's documentation for the specific configuration steps.

Using the MCP Server

Once connected, you interact with Common Room through natural language. The following examples show some of what's possible across any connected AI client.

Account Research

Get a full intelligence brief on any company in your workspace — including product signals, community engagement, intent data, CRM fields, and AI-generated summaries.

Example prompts:

"Research Datadog (datadog.com) for me"

"What's the latest signal activity for stripe.com?"

"Give me a quick overview of Notion — are they a good ICP fit?"

"What's the health score for my top 5 accounts?"

What you'll get: A structured brief covering the company snapshot, key contacts sorted by engagement score, recent product activity, community signals, open opportunities, and AI-sourced research. For targeted questions, the response focuses only on the relevant data.

Contact Research

Look up individuals by name, email address, LinkedIn handle, or GitHub username to understand their role, recent activity, and how to engage them.

Example prompts:

"Who is sarah.chen@acmecorp.com?"

"Look up Marcus Rivera at Stripe — is he active?"

"Tell me about the most engaged contacts at Datadog"

"Is @jsmith on GitHub associated with any accounts in my territory?"

What you'll get: A contact profile summary including their role and company context, recent product and community activity, website visit data, Spark enrichment and other activity.

Call Preparation

Before any customer or prospect meeting, get a structured brief that pulls together everything you need to know — account signals, attendee profiles, suggested talking points, and likely objections.

Example prompts:

"Prep me for my call with Figma tomorrow"

"I have a renewal QBR with Mixpanel next week — what should I know?"

"Prepare a brief for my meeting with Jane Doe (CPO) and Tom Smith (VP Eng) at Vercel"

"What are the top signals I should reference in my call with Datadog today?"

What you'll get: A meeting brief with the company snapshot, individual attendee profiles with persona classifications, the top signals to reference (product usage trends, community activity, intent data, recent news), 3–5 grounded talking points, 2–3 anticipated objections and how to handle them, and a recommended call outcome. If you've connected a calendar, Common Room can leverage the meeting context via the AI assistant.

Prospecting

Build targeted account or contact lists using signal data, firmographics, and segment membership. The MCP server distinguishes between your existing accounts (with full signal history) and net-new companies not yet in your workspace.

Example prompts:

"Find companies in my territory with high product usage scores that aren't in an open opportunity"

"Show me accounts in the Enterprise segment that have had declining engagement in the last 30 days"

"Find Series B fintech companies with 200–500 employees that I haven't touched yet"

"Who are the most engaged contacts at companies using our data integration features?"

"Build me a list of companies similar to my top 10 customers"

What you'll get: A filtered, sorted list of accounts or contacts with relevant signal data. For net-new companies (not yet in Common Room), results include firmographic data enriched with recent web signals like funding announcements and executive changes. You can refine criteria iteratively — "narrow that to companies with fewer than 500 employees" — and follow up with outreach drafting or account briefs for any result.

Outreach Composition

Generate personalized outreach messages grounded in real engagement signals via product activity, community contributions, website visits, job changes, and recent company news.

Example prompts:

"Draft an outreach email to Marcus Rivera at Stripe"

"Write a LinkedIn message to the top contact at Figma based on their recent activity"

"Compose outreach for all contacts at Acme Corp who visited our pricing page this week"

"I want to reach out to Jane Doe after her recent GitHub contributions — can you draft something?"

What you'll get: Personalized message for the contact, citing the specific signal it's based on, and copy which includes alternative personalization angles if you want to adjust the approach.

Available Tools

The Common Room MCP server exposes the following tools to connected AI clients. These tools are called automatically by the AI based on your natural language requests — you won't need to invoke them by name.

The commonroom_list_objects tool is the primary query interface. It supports:

• Object types: Contact, Organization, Activity, Segment, Tag, and more
• Filter operators: Filters on key properties filters
• Cross-object filtering: Filter contacts by organization attributes, organizations by contact attributes, or other custom attributes.
• Sorting and pagination: Sort by any supported field in ascending or descending order, with cursor-based pagination for large result sets

Tips for Best Results

Be specific about what you want. "Research Acme Corp" gives a full brief. "What's Acme Corp's product usage trend over the last 90 days?" gives a targeted answer to that specific question.

Mention your role for territory scoping. The MCP server retrieves your user profile automatically, but you can also say "show me accounts in my territory" or "what are my accounts with renewal risk" to scope results to your segments.

Combine workflows. After prospecting, ask for outreach drafts for the results. After a call brief, ask for a follow-up email. The AI maintains context across a conversation so you don't need to re-specify the account.

Use calendar integration for automatic prep. Connecting a Google Calendar or Outlook MCP server alongside Common Room enables automatic meeting detection for call prep and weekly briefings, saving you from listing meetings manually.

Iterate on prospect lists. Prospecting supports natural conversation — start broad, then refine: "narrow that to companies with more than 200 employees" or "show me only accounts with a fit score above 70."

Frequently Asked Questions

How does authentication work?

The Common Room MCP server uses OAuth 2.1. When you add the server URL to your AI client, the client will open a browser window redirecting you to Common Room's authorization page. After you log in and grant access, Common Room issues an OAuth token that your client stores and sends with each request. You won't need to handle tokens manually — the OAuth flow is handled entirely by your AI client and Common Room's auth server. Tokens are scoped to your user account, so the MCP server only returns data you already have permission to see in Common Room (respecting your workspace's role-based access controls).

Why can’t I complete the OAuth step when connecting Claude?

In some cases, a Claude admin must install the connection first before OAuth authorization can be completed in Step 3.

If the connection hasn’t been installed at the organization level, you may see issues or be unable to proceed with OAuth. In that situation, please reach out to your Claude administrator and ask them to install the connection. Once the installation is complete, you should be able to return and successfully complete the OAuth authorization step.

Can I disable MCP Server access to my instance?

Common Room administrators can disable MCP access for the entire instance by going to Settings → MCP Server in the Common Room app. That page also shows information about the users who have connected an MCP server, and the AI assistant used to connect.

The MCP server connected but I'm not seeing any data. What's wrong?

A few things to check: First, confirm you completed the OAuth flow fully — a partial authorization can result in a connected-but-unauthenticated state. Try disconnecting and reconnecting the server to re-trigger the auth flow. Second, make sure your Common Room account has access to the data you're querying (contacts, organizations, segments). Third, verify your workspace has data — if you're in a new or empty workspace, queries may return empty results legitimately.

My AI client says the tools aren't available. What should I check?

Most AI clients only expose MCP tools in specific modes. In GitHub Copilot, you must be in Agent mode — MCP tools are not available in Ask mode. In ChatGPT, you need to explicitly enable the Common Room connector in each new chat via the + menu. In Cursor, tools are available in the Agent panel but not in regular chat. Check your client's documentation to confirm you're in the right mode.

Can multiple team members connect to the same workspace?

Yes. Each person authenticates individually via OAuth using their own Common Room account. The MCP server respects each user's permissions — users will only see data their account has access to in Common Room. There's no shared token or service account required.

Why am I getting results for the wrong workspace?

If your Common Room account belongs to multiple workspaces, the MCP server uses your default workspace. You can explicitly tell your AI Assistant which Common Room instance to use (such as the RoomID).

Can the MCP server write data back to Common Room, or is it read-only?

The current MCP server is read-only. It supports querying contacts, organizations, activities, segments, and other objects, but does not write back to Common Room (no creating contacts, updating CRM fields, or logging activities). Outreach messages generated by the AI are drafted in your conversation only — sending them requires your own email or messaging tool.

The connection keeps dropping or I'm getting auth errors. How do I fix it?

OAuth tokens can expire or be revoked. Try disconnecting and reconnecting the MCP server to get a fresh token. If you recently changed your Common Room password or had your session revoked by an admin, you'll need to re-authorize. Some clients also cache a stale token — clearing your client's MCP server configuration and re-adding it from scratch typically resolves persistent auth errors.

I'm not finding a company or contact I know exists in Common Room. Why?

The MCP server queries your workspace based on your account's segment access. If the account or contact isn't in a segment you have access to, or if the search term doesn't match the name as stored in Common Room (e.g., a subsidiary vs. parent company name), it may not appear. Try querying by domain (e.g., acmecorp.com) rather than company name, or check directly in Common Room to confirm the record exists and is accessible to your user account.

Why aren’t custom connectors working inside Projects?

At this time, custom connectors are not supported within Projects. If you attempt to use an app or custom connector inside a Project, it may not function as expected.

This is a known limitation and has been reported by other users in the OpenAI community. You can follow the discussion and updates here:

https://community.openai.com/t/apps-custom-connectors-not-working-inside-projects/1369786

Why can’t I use MCPs in ChatGPT?

MCPs are not supported on Free plans. Access to MCP functionality is only available on certain paid plans.

If you’re currently using the Free plan, MCP features will not appear or may not function as expected.

You can review the full breakdown of app capabilities by plan in the “Apps capabilities by plan” section here:

https://help.openai.com/en/articles/11487775-apps-in-chatgpt