Connect your AI assistant to Perk

Connect an MCP-compatible AI assistant to Perk so you can ask questions about your company's travel, spend, invoice, and event data in natural language. Use this article to add the Perk MCP connector to your assistant and sign in with your Perk login.

Perk MCP works with AI assistants Claude and ChatGPT via native chat interface, and Gemini, VS Code and Cursor via CLI. Whichever assistant you use, you add the same Perk MCP server URL and sign in with your Perk credentials. Pick your assistant below.

Note: The Perk MCP is a remote server. Use the same URL everywhere: https://mcp.perk.com/api/mcp/mcp. Menu names in each assistant can change over time, so if a step looks different, check your assistant's own MCP documentation.

Connect Claude to Perk

Add Perk as a custom connector in Claude. 

  1. In Claude, go to Customize > Connectors.
  2. Search for Perk and click Connect.
  3. If you’re not already signed into Perk, it prompts you to sign in since the Perk MCP uses your existing Perk credentials through OAuth.
  4. In the Perk authorization window, click Authorize app.

If you aren’t already signed in, you may need to click Connect again on the Connectors page after signing into Perk.

Tip: If you can’t find Perk as a connector, check that it is available for your organization. Your Claude admin must first enable it for your organization:

  1. Go to Organization settings > Connectors.
  2. Select Add > Custom > Web.
  3. Enter Perk and the MCP URL: https://mcp.perk.com/api/mcp/mcp
  4. Click Save.

Connect ChatGPT to Perk

ChatGPT connects to Perk through developer mode, where an admin creates and publishes a custom MCP app. Full MCP is available on ChatGPT Business, Enterprise, and Edu; Pro users can connect it for read and fetch. 

Caution: Only a workspace admin or owner can turn on developer mode and publish the app. Custom MCPs can only be used on ChatGPT web, not mobile.

  1. An admin turns on developer mode in Settings > Apps > Advanced settings.
  2. In Workspace settings, go to Apps > Create.
  3. Enter the Perk MCP server URL https://mcp.perk.com/api/mcp/mcp with the required details, and choose OAuth for authentication.
  4. Click Scan Tools, then sign in with your Perk account when prompted and wait for the scan to finish.
  5. Click Create to save the app as a draft.
  6. An admin or owner publishes the app from Workspace settings > Apps > Drafts.

Connect Gemini to Perk

Connect Perk to Gemini through the Gemini CLI by adding the Perk MCP server to your settings file.

  1. Open or create ~/.gemini/settings.json.
  2. Add Perk under mcpServers:

    {
      "mcpServers": {
        "perk": {
          "httpUrl": "<https://mcp.perk.com/api/mcp/mcp>"
        }
      }
    }
    
  3. Restart the Gemini CLI.
  4. Sign in with your Perk account when prompted.

Connect VS Code to Perk

Add Perk to VS Code from the command line with the code --add-mcp command.

  1. In your terminal, run:

    code --add-mcp "{\"name\":\"perk\",\"type\":\"http\",\"url\":\"<https://mcp.perk.com/api/mcp/mcp\"}">
  2. Reload VS Code if prompted.
  3. Sign in with your Perk account when prompted.

Tip: You need the code command available in your terminal — in VS Code, run Shell Command: Install 'code' command in PATH from the Command Palette first.

Connect Cursor to Perk

Add Perk to Cursor's MCP settings. Use ~/.cursor/mcp.json to make Perk available in every project, or .cursor/mcp.json in a project folder to scope it to that project.

  1. Open or create ~/.cursor/mcp.json.
  2. Add Perk under mcpServers:

    {
      "mcpServers": {
        "perk": {
          "url": "<https://mcp.perk.com/api/mcp/mcp>"
        }
      }
    }
    
  3. Reload Cursor, then open Settings > MCP and check that Perk is connected.
  4. Sign in with your Perk account when prompted.

Connect from the command line

Coding agents that run in the terminal can add Perk with a single command. The examples below cover Claude Code and Codex.

For Claude Code:

claude mcp add --transport http perk <https://mcp.perk.com/api/mcp/mcp>

For Codex:

codex mcp add perk --url <https://mcp.perk.com/api/mcp/mcp>

After you add the server, start a session and sign in with your Perk account when prompted.

Allow report downloads

Perk MCP delivers travel and spend reports as a pre-signed download link, and your assistant fetches the file from Perk's cloud storage. For downloads to work, the owner of your AI assistant account needs to let the assistant reach these domains:

  • https://tk-prod-custom-report-download.s3.amazonaws.com — travel reports
  • http://storage.googleapis.com — spend reports

If these domains aren't on your assistant's allowlist, a report can generate but the download won't open.

Things to know

Keep the following in mind after you connect your assistant to Perk:

  • You can query travel and spend data and generate reports, but you can't book, change, or send anything through your assistant.
  • Your data is scoped to your Perk role and your company's modules. You only see data you could already see in Perk.
  • Write actions, such as booking and sending expenses, are coming soon in a future release.

Note: You authenticate with your existing Perk credentials through OAuth. The assistant acts within your existing permissions and can't access data you couldn't already see in Perk.

Where to go next

Now that your assistant is connected, try a first prompt. For guidance, see Ask Perk your first question. If your assistant can't connect or you see less data than expected, see Troubleshoot Perk MCP.

Was this article helpful?