Connect your local AI assistant

The MCP server lets an MCP-compatible AI client discover and use your application operations: upload and organize files, apply labels and categories, manage cases and hearings, create documents, collaborate, and export. Your client only sees tools allowed by its token, and the server checks permissions again when each tool runs.

1. Create a dedicated token

Open API tokens. Start with read access if you want to inspect data; add write permissions for changes. Choose full access only when your assistant needs all operations. Choose an expiration date or no expiration. Keep this token private; local clients and their model providers may receive data returned by the tools.

2. Connect using Streamable HTTP

In a client that supports remote MCP with custom headers, configure the following URL and header. Client configuration keys vary; use your client’s remote HTTP connection settings.

URL: https://app.courtbriefly.com/api/mcp
Transport: Streamable HTTP
Authorization: Bearer YOUR_API_TOKEN

This server uses personal-token authentication, not an OAuth sign-in flow. It supports stateless Streamable HTTP with JSON responses through the official MCP TypeScript SDK. The client negotiates the supported protocol version. Use the stdio adapter below if your client cannot supply an Authorization header.

3. Or use the local stdio adapter

Install Node.js 20 or newer. Download the adapter and save it to an absolute path on your computer. Add this entry to the MCP server configuration in a stdio-compatible client:

{
  "mcpServers": {
    "case-commander": {
      "command": "node",
      "args": [
        "/absolute/path/case-commander-mcp.mjs"
      ],
      "env": {
        "CASE_COMMANDER_URL": "https://app.courtbriefly.com",
        "CASE_COMMANDER_API_TOKEN": "YOUR_API_TOKEN"
      }
    }
  }
}

Replace the path and token, save, and restart or reconnect the MCP server in your client. Keep the configuration out of shared repositories. The adapter requires no npm packages and forwards the same authenticated protocol to the server. It never writes token values to logs. Clients using TOML or another format should use the same command, arguments, and environment values.

Try a workflow

Your AI client also needs access to the local file. The hosted MCP server cannot read your computer’s filesystem. Upload tools accept a files array with field, name, mimeType, and base64 bytes. Ordinary fields go in body. Files are limited to 16 MiB each and MCP request bodies to 24 MiB; use multipart REST uploads for larger files within your deployment’s request limit.

{
  "name": "post_evidence",
  "arguments": {
    "body": {
      "caseId": "YOUR_CASE_ID",
      "folderId": "YOUR_FOLDER_ID"
    },
    "files": [
      {
        "field": "files",
        "name": "receipt.txt",
        "mimeType": "text/plain",
        "base64": "U2Nob29sIHJlY2VpcHQ="
      }
    ]
  }
}

Dynamic IDs go in path, query filters in query, and JSON changes in body. Use context: "ws:WORKSPACE_ID" or "matter:CASE_ID" to select a context. Binary downloads return base64 and MIME type, up to 16 MiB; use REST for larger downloads.

Troubleshooting and usage

• No tools: check the token’s permissions, expiration, revocation status, and email verification.
• 401: update the Authorization header or the adapter’s token environment variable.
• 403: check both token scope and the account’s sharing/role permissions. Reconnect after changing scopes to refresh the visible tools.
• 402: an existing subscription, storage, or AI allowance has been reached. Connecting a local AI does not bypass these limits.
• 429: wait for Retry-After. API and MCP share the account’s request budget.
• Local AI reasoning does not spend app AI operations. Explicitly invoking built-in AI tools does, under the same rules as the website.