Reka MCP Server
Upload, index, and analyze video content using the Reka Vision API and natural language.
Data last scanned today · Reviewed today
Overview
The Reka MCP Server provides tools to bridge AI agents with the Reka Vision API. It enables users to upload, index, and search video libraries while performing deep visual analysis, such as object detection within specific time ranges and transcript extraction. The server supports both local stdio execution and hosted HTTP deployments. In hosted mode, the server acts as an interface that forwards API keys on a per-request basis, facilitating integration with MCP clients that support streamable HTTP.
Our verdict
This server carries a high risk level with a security score of 24/100, primarily due to it being a community-maintained project rather than an official vendor release. With no editorially reviewed authentication methods and a lack of license specification, users should exercise caution when deploying it in environments handling sensitive data or credentials.
- Supports direct natural language querying of video content
- Provides both local stdio and hosted HTTP transport modes
- Features a minimal implementation with zero direct dependencies
- Lacks an official license, posing potential usage uncertainties
- Security architecture has not yet undergone editorial review
- Maintained by a single contributor with low repository activity
Tools
| Tool | Description | Risk |
|---|---|---|
| upload_video | Upload a video from a URL | medium |
| list_videos | List videos in your account or a group | low |
| get_video | Get video details, metadata, and feature status | low |
| update_video | Update a video's name, title, description, or group | medium |
| delete_video | Permanently delete a video and all indexed data | high |
| create_group | Create a new video group | medium |
| list_groups | List all video groups | low |
| delete_group | Delete a video group | high |
| index_video | Index a video for search/QA/analysis. Waits until all requested features are ready (2-10 min). | medium |
| search_videos | Semantic search across indexed videos | low |
Compatibility
| Client | Local | Docker | Remote | Read-only |
|---|---|---|---|---|
| Claude Desktop | ||||
| Cursor | ||||
| VS Code | ||||
| Windsurf | ||||
| ChatGPT |
Frequently asked questions
›What is the difference between local mode and hosted mode?
In local mode, the Reka API key is set once via the environment variable for all requests. In hosted mode, each MCP HTTP request must provide the API key in the X-Reka-API-Key header, which is then forwarded to the Reka Vision API.
›How do I correctly index a video for analysis?
First, use upload_video, then poll get_video until its status is listed as uploaded. Once uploaded, call index_video, which will poll until all requested features are ready.
›Can I connect Claude Desktop to the hosted version of this MCP server?
Yes, but because Claude Desktop does not support Streamable HTTP natively, you must use the mcp-remote bridge to connect to the hosted URL.
›What should I do if I need to update the server to the latest version?
Run uv cache clean reka-mcp to clear the cached package and then restart your MCP client.
›How can I secure my hosted server with transport-level authentication?
You can set the REKA_MCP_AUTH_TOKEN environment variable, which requires HTTP clients to include an Authorization: Bearer token in their requests.
›How do I perform local testing for hosted mode?
You must override the default DNS rebinding protection by setting REKA_MCP_ALLOWED_HOSTS and REKA_MCP_ALLOWED_ORIGINS to include your local address, such as localhost or 127.0.0.1.
Badge
Maintain this server? Add the live badge to your README.