Use DeeplyClear inside your AI tools

What Your Agent Can See

Map tools return structured nodes and edges, an inline PNG preview image, and a live DeeplyClear map link. ChatGPT can render the interactive canvas widget when the DeeplyClear app is connected; other MCP clients can use the PNG preview and link.

The read, create, prompt-edit, elaborate-node, and operation tools include previews automatically. Agents can also call `deeplyclear_preview_map` when they only need a fresh visual snapshot.

1. 1In ChatGPT, open Settings, then Apps & Connectors.
2. 2Enable developer mode if you are testing before the DeeplyClear app is published.
3. 3Create a connector and paste https://mcp.deeplyclear.com. If ChatGPT asks for an explicit MCP path, use https://mcp.deeplyclear.com/mcp.
4. 4When ChatGPT asks to connect, sign in to DeeplyClear and approve the requested workspace access.

ChatGPT discovers DeeplyClear OAuth from the MCP server metadata and uses an authorization-code flow with PKCE. You do not need to paste a manual token into ChatGPT.

In developer-mode testing, reconnect the draft app after MCP changes so ChatGPT refreshes the tool metadata and loads the DeeplyClear canvas widget.

1. 1Add the MCP server to your global ~/.codex/config.toml or a trusted project .codex/config.toml.
2. 2Run codex mcp login deeplyclear and approve the browser-based DeeplyClear OAuth screen.
3. 3Restart Codex or open /mcp in the Codex TUI to confirm the server is loaded.

[mcp_servers.deeplyclear]
url = "https://mcp.deeplyclear.com"

If you need non-interactive setup, create a DeeplyClear token from AI tool connections, set `DEEPLYCLEAR_MCP_TOKEN`, and use bearer-token config instead.

[mcp_servers.deeplyclear]
url = "https://mcp.deeplyclear.com"
bearer_token_env_var = "DEEPLYCLEAR_MCP_TOKEN"

1. 1Add DeeplyClear as an HTTP MCP server.
2. 2Run claude mcp login deeplyclear, or open /mcp inside Claude Code and authenticate from there.
3. 3Approve the browser-based DeeplyClear OAuth screen.

claude mcp add --transport http deeplyclear https://mcp.deeplyclear.com
claude mcp login deeplyclear

For team-shared or non-interactive configuration, use a bearer token in an environment variable with Claude Code's `.mcp.json` expansion.

claude mcp add --transport http deeplyclear https://mcp.deeplyclear.com \
  --header "Authorization: Bearer YOUR_DEEPLYCLEAR_MCP_TOKEN"

{
  "mcpServers": {
    "deeplyclear": {
      "type": "http",
      "url": "https://mcp.deeplyclear.com",
      "headers": {
        "Authorization": "Bearer ${DEEPLYCLEAR_MCP_TOKEN}"
      }
    }
  }
}

1. 1Open Cursor settings and find the MCP or Tools & Integrations section.
2. 2Create or edit your mcp.json file.
3. 3Paste a remote MCP server entry for DeeplyClear, then restart or reconnect MCP tools.

{
  "mcpServers": {
    "deeplyclear": {
      "url": "https://mcp.deeplyclear.com",
      "headers": {
        "Authorization": "Bearer YOUR_DEEPLYCLEAR_MCP_TOKEN"
      }
    }
  }
}

Store this in a user-level Cursor config when possible. If you keep MCP config in a repository, do not commit a real token.

Other MCP Clients

Use Streamable HTTP, the hosted endpoint, and an `Authorization` header:

Transport: Streamable HTTP
URL: https://mcp.deeplyclear.com
Authorization: Bearer YOUR_DEEPLYCLEAR_MCP_TOKEN

After connecting, ask the agent to list your DeeplyClear maps or create a new map from a prompt. Access is limited to the workspace that created the token.