How to connect

Values used everywhere

Each example below comes in two forms: connecting by logging in (no key written) and with a key. Without a key written, the login screen opens the first time you use it.

Claude Code

Connect by logging in (no key written; the login screen opens the first time you use it):

claude mcp add --transport http latentdatabank https://latentdatabank.com/mcp?lang=en

With a key:

claude mcp add --transport http latentdatabank https://latentdatabank.com/mcp?lang=en --header "Authorization: Bearer YOUR_KEY"

To share it inside a project, write it in .mcp.json at the project root. The key is read from an environment variable (LDB_API_KEY).

{
  "mcpServers": {
    "latentdatabank": {
      "type": "http",
      "url": "https://latentdatabank.com/mcp?lang=en",
      "headers": {
        "Authorization": "Bearer ${LDB_API_KEY}"
      }
    }
  }
}

Claude (connectors on the web and in the desktop app)

1. Open Customize, then Connectors, and choose + Add, then Add custom connector.
2. Enter a name (for example Final Check) and the MCP URL https://latentdatabank.com/mcp?lang=en, then press Continue.
3. Press Add. The connect screen opens; log in with a company account and press Continue (no key needed).
4. In a chat, turn it on from the + button, then Connectors.

On Team and Enterprise plans an owner adds it under Organization settings, then Connectors. Claude connects to our MCP server from Anthropic's servers.

Claude Desktop (config file) and clients that launch a command

Clients that take a command in a config file (claude_desktop_config.json) connect through mcp-remote, a published bridge (needs Node 18 or later).

Connect by logging in (no key written; the login screen opens the first time you use it):

{
  "mcpServers": {
    "latentdatabank": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://latentdatabank.com/mcp?lang=en"
      ]
    }
  }
}

With a key (Bearer and the key go in an environment variable so that the space is not lost):

{
  "mcpServers": {
    "latentdatabank": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://latentdatabank.com/mcp?lang=en",
        "--header",
        "Authorization:${LDB_AUTH}"
      ],
      "env": {
        "LDB_AUTH": "Bearer YOUR_KEY"
      }
    }
  }
}

Cursor

Write it in .cursor/mcp.json for a project, or ~/.cursor/mcp.json for all projects.

Connect by logging in (no key written; the login screen opens the first time you use it):

{
  "mcpServers": {
    "latentdatabank": {
      "url": "https://latentdatabank.com/mcp?lang=en"
    }
  }
}

With a key (read from the environment variable LDB_API_KEY):

{
  "mcpServers": {
    "latentdatabank": {
      "url": "https://latentdatabank.com/mcp?lang=en",
      "headers": {
        "Authorization": "Bearer ${env:LDB_API_KEY}"
      }
    }
  }
}

ChatGPT

1. Open chatgpt.com/plugins, select the plus button, then Add custom MCP server.
2. Enter a name and a description.
3. Under Connection, enter https://latentdatabank.com/mcp?lang=en as the public endpoint.
4. Choose OAuth (log in) for authentication. Review the risk warning and select I understand and want to continue, then Create as a plugin; the connect screen opens. Log in with a company account, press Continue, and review the tools that were found.

Once you connect by logging in, every tool is found. An order is placed only after the user has seen the order sheet and clearly said yes.

OpenAI API (Responses)

With a key (put the key in authorization; the API cannot open the login screen, so use a key):

{
  "model": "your-model",
  "tools": [
    {
      "type": "mcp",
      "server_label": "latentdatabank",
      "server_url": "https://latentdatabank.com/mcp?lang=en",
      "authorization": "YOUR_KEY"
    }
  ],
  "input": "Please have someone call the hair salon by the station and ask for free slots on Saturday afternoon. My cap is 3000 yen"
}

Gemini CLI

Connect by logging in (no key written; the login screen opens the first time you use it):

gemini mcp add --transport http latentdatabank https://latentdatabank.com/mcp?lang=en

With a key:

gemini mcp add --transport http --header "Authorization: Bearer YOUR_KEY" latentdatabank https://latentdatabank.com/mcp?lang=en

Or in the settings file (~/.gemini/settings.json, or .gemini/settings.json in a project):

{
  "mcpServers": {
    "latentdatabank": {
      "httpUrl": "https://latentdatabank.com/mcp?lang=en",
      "headers": {
        "Authorization": "Bearer YOUR_KEY"
      }
    }
  }
}

Other HTTP MCP clients

Enter https://latentdatabank.com/mcp where the client asks for a Streamable HTTP URL. Clients that support OAuth (login) open the connect screen; log in with a company account and press Continue. For other clients, add the header Authorization: Bearer KEY. With neither a key nor a login, the answer is 401 with WWW-Authenticate (the URL that points to the login). This one line checks a connection with a key (the list of tools comes back as JSON).

curl -X POST -H "Content-Type: application/json" -H "Authorization: Bearer YOUR_KEY" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' \
  "https://latentdatabank.com/mcp?lang=en"

REST (curl)

Categories (no key):

curl "https://latentdatabank.com/api/agent/categories?lang=en"

Guide price (no key):

curl -X POST -H "Content-Type: application/json" \
  -d '{"category":"call","due_within":"1h"}' \
  "https://latentdatabank.com/api/agent/estimates?lang=en"

Build an order sheet (no key; the only things to ask the person are what they want done and the budget cap, and everything else is filled in automatically; if either is missing, it comes back as a question):

curl -X POST -H "Content-Type: application/json" \
  -d '{"request":"Call the shop and ask for free slots today","category":"call","fields":{"due_within":"1h","budget_max_yen":3000}}' \
  "https://latentdatabank.com/api/agent/orders?lang=en"

With a key (search workers):

curl -H "Authorization: Bearer YOUR_KEY" "https://latentdatabank.com/api/agent/workers?category=call&limit=5"

The list of entry points and the shape of the answers are on Asking from an AI agent.

Create a key

1. Log in with a company account (or sign up; it is free).
2. On "AI agent keys" (agentkey), choose a name and a type (live or test) and create the key. Test keys return sample data.
3. The full key is shown only once, right after you create it. Copy it into the settings above.

Steps, prices and caps, where a person must confirm and what cannot be done are on Asking from an AI agent. How to ask for each kind of task:

• How to have a phone call made in Japan from an AI agent
• How to have a place in Japan checked and photographed from an AI agent
• How to have shopping done in Japan from an AI agent
• How to have a person review what AI made
• How to have a person in Japan try an app or a shop and give feedback, from an AI agent