Connect Claude, ChatGPT, or Cursor to In Parallel so your AI assistant can read your meetings, goals, and Workspaces — and act on them — directly from a chat.

In Parallel exposes its data and actions through the Model Context Protocol (MCP), the open standard AI clients use to talk to external tools. Once connected, your AI tool can look up your meeting records, list decisions and action items, run an execution-health check, and more — all within the access you grant.

Note: Claude, ChatGPT, and Cursor are supported today. GitHub Copilot is coming next.

Both walkthroughs below finish at the same place: an In Parallel OAuth consent screen where you choose which Workspaces to share and click Allow access. If you've connected an AI tool before, that step will look familiar.

These steps use the Claude desktop app. The connector you add is a custom MCP connector pointing at In Parallel.

In the Claude desktop app, open your account menu and choose Settings.

In Settings, open the Connectors section in the left nav. Claude shows a notice that "Connectors have moved to Customize. Head there to browse, connect, and manage them." Click the Customize link to continue.

In Customize → Connectors, find In Parallel under Web connectors and select it. If it isn't listed yet, use the + (add) icon to add a connector, enter the In Parallel MCP URL — https://www.in-parallel.ai/mcp — and then select it.

In Parallel opens its connect page showing the MCP URL https://www.in-parallel.ai/mcp and "You are not connected to In Parallel yet." Click Connect to start authorization.

The In Parallel consent screen appears. It shows what the AI tool will be able to do — access your In Parallel account and data and act on your behalf via the Model Context Protocol — and asks you to select which Workspaces to grant access to:

Review the selection, then click Allow access. Authorizing redirects you back to your AI client.

Back in Claude, In Parallel shows as connected. Under Tool permissions you can choose when Claude is allowed to use each tool. Write and delete tools default to Needs approval, and you can switch a group or an individual tool to Needs approval, Blocked, or Custom.

In any chat, open the + menu, choose Connectors, and toggle In Parallel on. Claude can now use In Parallel tools in that conversation.

In ChatGPT, In Parallel is added as a custom app (MCP server). Custom MCP apps are part of ChatGPT's Developer Mode, which is currently in beta — so the app is created under Advanced settings, and chats that use it are labelled "Developer mode."

Note: This is ChatGPT's beta path for custom MCP servers; the exact wording and location may change as OpenAI moves it out of beta. Cursor and Claude have their own connector UIs (above / below).

Open your ChatGPT account menu, choose Settings, then open the Apps section. This is where you manage enabled apps ChatGPT can use in your chats.

On the Apps screen, under Advanced settings, click Create app to open the New App form (marked beta). Fill it in:

ChatGPT shows a warning that custom MCP servers introduce risk; acknowledge it with the I understand and want to continue checkbox, then click Create.

The same In Parallel consent screen appears. Choose Full access or select individual Workspaces, then click Allow access. Authorizing redirects you back to ChatGPT.

In a chat, open the + menu, go to More, and choose In Parallel to enable the connector for that conversation. Because In Parallel is a custom MCP app, the chat runs in Developer mode (shown at the top of the composer) and memory is not used for that chat.

Once enabled, the In Parallel connector chip shows below your message, and you can ask it about your In Parallel data.

Cursor is supported. Add In Parallel as an MCP server in Cursor's settings, using the same MCP URL — https://www.in-parallel.ai/mcp — and complete the OAuth consent step described above (choose your Workspaces, then Allow access). Cursor's MCP configuration changes occasionally as the client evolves, so follow Cursor's own MCP setup docs for the exact location of the server config.

GitHub Copilot is coming next — it isn't supported yet.

Once connected, your AI tool can read from In Parallel — your meetings, goals, decisions, action items, and Workspaces — and run actions like creating an action item or running the get_drift_report execution-health diagnostic. Each In Parallel tool has a clear, human-readable title so your AI client picks the right one. For the full set, see the In Parallel MCP tools reference.

You stay in control of what it can do: per-tool permissions let you require approval before any write or delete tool runs, or block specific tools entirely (see step 6 of the Claude walkthrough). Write and delete tools require approval by default.