DEPRECATED: The local stdio MCP server is deprecated. Use the remote MCP server at https://mcp.echosaw.com instead. See MCP Integration Guide for setup instructions. The remote server uses OAuth 2.0 and requires no local installation.

The one thing the remote server cannot do is upload a file from your local disk: echosaw_analyze_media exists only here. Keep this package installed if you need local-file submission; use echosaw_analyze_media_url on the remote server otherwise.

@echosaw/mcp-server

Echosaw MCP Server — AI assistant integration for media intelligence.

The Echosaw MCP server connects AI assistants (Claude Desktop, Cursor, and other MCP-compatible clients) to Echosaw's media intelligence platform. Submit media for analysis, check status, retrieve intelligence reports, and download source files — all from within your AI-powered development environment.

Prerequisites

• Node.js 18+ (minimum version supported by @modelcontextprotocol/sdk)
• Echosaw API key — Your API key is e-mailed to you after subscribing to a Starter, Growth, Pro, or Agency plan.

Installation

npm install -g @echosaw/mcp-server

Environment Variables

| Variable | Required | Default | Description | |----------|----------|---------|-------------| | ECHOSAW_API_KEY | Yes | — | Your Echosaw API key | | ECHOSAW_API_URL | No | https://api.echosaw.com | API base URL (override for gamma/staging) |

Client Configuration

Claude Desktop

Add the following to your Claude Desktop configuration file (claude_desktop_config.json):

{
  "mcpServers": {
    "echosaw": {
      "command": "echosaw-mcp",
      "env": {
        "ECHOSAW_API_KEY": "your-api-key"
      }
    }
  }
}

Cursor

Add to your Cursor MCP settings:

{
  "mcpServers": {
    "echosaw": {
      "command": "echosaw-mcp",
      "env": {
        "ECHOSAW_API_KEY": "your-api-key"
      }
    }
  }
}

Available Tools

| Tool | Description | |------|-------------| | echosaw_analyze_media | Start a new analysis job by uploading a local audio, video, or image file from disk. Entry point that creates a job and begins processing (does not fetch previously analyzed media). Returns a job ID used to track processing and retrieve results. | | echosaw_analyze_media_url | Start a new analysis job from a publicly accessible or authorized media URL without uploading the file directly. Entry point that creates a job and begins processing (does not fetch previously analyzed media). Returns a job ID used to track processing and retrieve results. | | echosaw_check_job_status | Retrieve the current processing state of an Echosaw analysis job, including whether the job is queued, processing, completed, or failed. | | echosaw_get_analysis_results | Retrieve structured analysis results generated by Echosaw for a completed job, including summaries, transcripts, detected entities, events, and other intelligence outputs. Supports an optional section parameter (summary, transcript, safety, insights, metadata, downloads, all) to reduce response size. | | echosaw_download_media | Retrieve the original source media file of an already-analyzed job by generating a presigned download URL for it. Fetches existing media only — does not upload or start analysis. The URL is valid for 1 hour. |

The remote MCP server at https://mcp.echosaw.com includes additional tools: echosaw_search_media, echosaw_list_media (supports server-side filters for mediaType (VIDEO/AUDIO/IMAGE), job status (e.g. COMPLETE), report sentiment (positive/negative/neutral/mixed), and Public Library visibility (isPublic: true/false) in addition to limit and folderPath), echosaw_get_profile, echosaw_get_organization (retrieve your Echosaw organization details — name, your role, creation date, and active member count; returns null if you are not a member of any organization), echosaw_list_folders, echosaw_create_folder, echosaw_move_media, echosaw_rename_folder, echosaw_delete_folder, echosaw_set_media_visibility (toggle Public Library visibility for a media item; requires COMPLETE status; blocked with moderation flags returned when content fails safety checks).

Response Format

All tools return a standardized JSON response:

| Field | Type | Description | |-------|------|-------------| | success | boolean | Whether the operation succeeded | | tool | string | The tool that produced this response | | responseVersion | string | Response schema version (currently "1.0") | | data | object | Tool-specific result data | | message | string? | Human-readable summary | | error | string? | Error description (when success is false) | | nextAction | object? | Suggested next tool to call |

Example Workflow

1. Submit a file for analysis:

   "Analyze the video at /Users/me/Downloads/interview.mp4"

   The assistant calls echosaw_analyze_media with the path and mediaType: "video", returning a mediaId.

2. Check processing status:

   "What's the status of that analysis?"

   The assistant calls echosaw_check_job_status with the mediaId.

3. Retrieve results:

   "Show me the results"

   Once complete, the assistant calls echosaw_get_analysis_results to retrieve the full intelligence report including transcript, safety analysis, content insights, and metadata.

4. Download source media:

   "Get me a download link for that video"

   The assistant calls echosaw_download_media to get a presigned URL valid for 1 hour.

Development

# Clone and build
cd lib/mcp-server
npm install
npm run build

# Run directly
ECHOSAW_API_KEY=your-key node dist/bin/echosaw-mcp.js