Skip to main content
Connect Deepline to a client that supports remote MCP servers and OAuth sign-in. A client without a shell needs no CLI installation: the client connects directly to Deepline’s hosted server. For a hosted agent, paste this into its chat to set up the connection:
The agent gives you the steps for your client. You add the connection, complete sign-in and workspace authorization, and enable Deepline in your conversation. The agent then verifies it. The agent setup guide works across hosted clients that support custom remote MCP connections, without a directory listing or local installation.

1. Add the remote connection

Expand your client below. Complete the Deepline authorization in step 2 when the client opens sign-in, then return to finish adding or enabling the connection.
  1. Open ChatGPT Plugins, select the + button, then Add custom MCP server.
  2. Name it Deepline. Under Connection, enter https://code.deepline.com/api/v2/mcp as the server URL.
  3. Choose OAuth. Use automatic client registration (DCR) if asked; leave the client ID and secret empty.
  4. Review the connection notice, select I understand and want to continue, then Create as a plugin. Install the resulting plugin from your personal plugins or workspace and complete sign-in.
  5. In a new conversation, type @ and select Deepline before sending the discovery prompt below.
Workspace permissions can restrict adding or using custom servers. See OpenAI’s custom MCP instructions if the option is unavailable.
The remote connector belongs to your Claude account and works across Claude web, Desktop, and Cowork.
  1. Open Customize > Connectors, then + Add > Add custom connector.
  2. Name it Deepline, enter https://code.deepline.com/api/v2/mcp, and click Continue.
  3. Keep OAuth authentication. Choose Sign in now and Register automatically for the OAuth client, then finish with Add.
  4. Complete Deepline sign-in. In your conversation, open + > Connectors and enable Deepline.
For Team or Enterprise, an authorized administrator first adds it under Organization settings > Connectors > Add > Custom > Web. Members then find Deepline under Customize > Connectors and click Connect. See Claude’s remote-connector instructions.For Deepline skills and CLI execution in Cowork, use the Claude plugin.
Custom apps currently require a personal Google Account, age 18 or over, US availability, English, and Keep Activity enabled. Work and school accounts are not supported.
  1. Open Gemini, then Settings > Connected Apps. If needed, go through Personal Intelligence > Connected Apps.
  2. Under Custom apps, choose Add a custom app or the field inviting you to add a custom app link.
  3. Enter https://code.deepline.com/api/v2/mcp and click Next. Use automatic registration; leave advanced client credentials empty.
  4. Complete Deepline sign-in. In your chat, type @, select the Deepline custom app, and send the discovery prompt below.
Add the connection on the web; Google also makes it available in Gemini’s mobile app. See Google’s custom-app instructions for current availability.
Custom remote connectors are available on Pro, Max, and Enterprise. Your organization’s administrator may need to allow member-added connectors.
  1. Open Plugins in the sidebar, or Customize > Connectors if Plugins is not shown. Choose Add custom > Add MCP connector.
  2. Select Remote, name it Deepline, and enter https://code.deepline.com/api/v2/mcp as the MCP server URL.
  3. Choose OAuth authentication and Streamable HTTP transport. Leave client credentials empty for automatic registration.
  4. Review the acknowledgement and click Add. Open the connector card to authenticate and enable it.
  5. In your conversation or Computer workflow, explicitly ask to use the Deepline connector with the discovery prompt below.
See Perplexity’s remote-connector instructions.
  1. Open .cursor/mcp.json in your project, or ~/.cursor/mcp.json for all projects. Add the deepline entry alongside any existing servers:
  1. Save the file, open Customize > MCPs, and enable Deepline. Follow Cursor’s OAuth sign-in prompt; no API-key header is needed.
  2. Return to an Agent conversation and send the discovery prompt below.
See Cursor’s MCP configuration. For an agent that should execute Deepline through its terminal, use CLI installation.
  1. Open Cursor Agents and its MCP dropdown. Add a personal custom server using HTTP and https://code.deepline.com/api/v2/mcp.
  2. Name it Deepline and complete OAuth sign-in. Each user authorizes their own connection, including for team-shared servers.
  3. Enable Deepline in the MCP dropdown, then send the discovery prompt below in the agent session.
Team administrators can add shared servers under Dashboard > Plugins & MCPs. A local editor configuration alone does not configure a cloud agent. See Cursor’s cloud-agent MCP instructions.
  1. Open a project in VS Code with Copilot enabled. In the Command Palette, run MCP: Add Server.
  2. Choose HTTP, enter https://code.deepline.com/api/v2/mcp, and use deepline as the server ID.
  3. For a personal connection, choose Global > Copilot Global when prompted. Review the server trust prompt and complete OAuth sign-in.
  4. Open a Copilot agent session. Enable Deepline’s tools under Chat: Open Customizations > Tools, then send the discovery prompt below.
These steps apply to Copilot in VS Code. Configuration destinations depend on the selected agent harness; see VS Code’s MCP setup instructions.
Open your client’s remote-server or custom-connector settings and enter:Use automatic OAuth client registration if offered. You do not need to create your own OAuth app or paste a provider API key into these fields. Complete Deepline sign-in, enable the connection, and send the discovery prompt below.The client must support custom remote MCP servers with OAuth; choosing a model or pasting the URL into a message does not connect it.

2. Sign in and authorize the workspace

Complete Deepline sign-in yourself in the authorization window. Your Deepline email must be verified before MCP authorization can complete. On Authorize MCP Access, check the client name, workspace, requested access, and redirect URI. If you belong to multiple workspaces, select the one this connection should use. Click Authorize, then return to your MCP client. The connection uses that workspace’s permissions and applicable Deepline usage.

3. Enable the connection and verify discovery

Enable Deepline in the conversation using your client’s instructions above. Ask the client to inspect the connection without starting a lookup:
Inspect the client’s tool calls. Confirm that a Deepline tool returned the workflow identifier and declared inputs; a plausible answer alone does not verify the connection. Catalog search and contract lookup read metadata; they do not run a Play or spend Deepline credits. A search result does not mean every provider action is authorized. If the tools do not display the workspace, confirm it from the authorization screen rather than asking the client to guess.

If the connection fails

Check the failing step before creating another connection. For an authorization error, verify your Deepline email, finish sign-in, and confirm the workspace on the consent screen. If the connection uses the wrong workspace or its authorization is no longer valid, reconnect and authorize the intended workspace. If no tools appear, confirm Deepline is enabled in the current conversation.
For a network error, check the exact server URL and the client’s access to it. Claude’s remote connection originates from Anthropic’s servers, so reaching an endpoint from your own browser does not prove Claude can reach it. A provider-access error after connection is a separate issue from MCP sign-in or tool discovery; keep its error evidence when choosing the next action.

Continue with a task

Setup stops at discovery. Use the quickstart or work-email guide for a reviewed first lookup. For another installation route, use CLI setup or the Claude plugin for Claude Code and Cowork.