Model Context Protocol

In order to use the logs and data you collect in RunReveal with Cursor, Claude, or other LLMs, you can use RunReveal’s CLI support of MCP.

This works by having the runreveal cli as a broker between the LLM and RunReveal’s API, handling authentication, authorization, and the actual fetching of data.

We’ll likely continue to expand the capabilities of LLMs in RunReveal but the space is changing quickly and we’re on the bleeding edge.

Install the CLI

Consult the RunReveal CLI docs for installation instructions through brew.

Run which runreveal, take note of the full path to the binary. If you used brew it’s likely in /opt/homebrew/bin/runreveal.

Get a RunReveal API Key and your Workspace ID

To use the MCP server you’ll need to configure the RunReveal CLI via an API token. The RunReveal CLI will store your credentials in the keychain, which currently has some integration issues with the LLM Clients.

To issue an API key go to Settings > API Keys and create a new API key with Analyst permissions. This allows the key to search logs but won’t allow it to delete or create additional resources.

Additionally, make note of your workspace ID on the Settings page, it should be a long string that looks something like this 2KUOdhvRReF5RZg31235lfq3fde

Connect Claude to RunReveal

Inside Claude Desktop, under Settings > Developer > Edit MCP config, it should open a file that can manually be edited. This file is likely in vim ~/Library/Application\ Support/Claude/claude_desktop_config.json on a Mac.

Inside this file you’ll need to configure your Claude client to understand how to talk to the RunReveal binary. This requires a RunReveal API key, and a RunReveal Workspace ID.

If you don’t have an MCP server set up already, your config file should look like this with the RUNREVEAL_CONFIG filled in. Otherwise, insert a new entry in the mcpServer object for runreveal.

{
    "mcpServers": {
        "runreveal": {
            "command": "/opt/homebrew/bin/runreveal",
            "args": [
                "mcp"
            ],
            "env": {
                "RUNREVEAL_TOKEN": "YOUR_API_KEY_FROM_RUNREVEAL",
                "RUNREVEAL_WORKSPACE": "YOUR_WORKSPACE_ID"
            }
        }
    }
}

Once you save this file you should be able to open Claude desktop and see RunReveal under Developer Settings listed as a MCP Server.

Consult the Claude docs for additional details.

Connect Cursor to RunReveal

To use Cursor along with RunReveal you’ll need to provide the same configuration in your ~/.cursor/mcp.json file in your home directory or project directory. The format of this file, currently, is in the same format as the Claude client format.

{
    "mcpServers": {
        "runreveal": {
            "command": "/opt/homebrew/bin/runreveal",
            "args": [
                "mcp"
            ],
            "env": {
                "RUNREVEAL_TOKEN": "YOUR_API_KEY_FROM_RUNREVEAL",
                "RUNREVEAL_WORKSPACE": "YOUR_WORKSPACE_ID"
            }
        }
    }
}

Consult the cursor docs for additional details. Once you restart cursor you should see the RunReveal mcp server listed in Cursor settings. Make sure you don’t configure the MCP server through the Cursor UI, because this won’t allow you to configure necessary envvars.