jevreel
Find and verify moments in video using contact sheets and a decision model, no embeddings
- Pricing
- Open source
- Platforms
- CLI, macOS, Linux
- Runs
- Locally
- Version
- 0.2.0
- Listing updated
- today
Similar MCP servers
About jevreel
jevreel is an open-source MCP server for design and media. It runs locally on your machine and is started by your AI client.
jevreel lays a video out on labelled contact sheets of 60 frames each, sampled at one frame per second, and has Jev, TypeSafe's decision model, judge every sheet in parallel. The best windows are rechecked on full frames every half second, and the agent gets back moments, times and probabilities, then looks closer itself. No vectors are computed or stored.
Because questions are asked at search time, the same index can answer a yes/no with the second it happens, pick one of your own options, or score a level, per minute or for the whole film. Custom lenses add typed questions to every minute at little extra cost, and the README shows searching dialogue from captions, reading on-screen text and finding a person from a reference photo. Answers are cached.
It needs Python 3.11 or newer, FFmpeg and a TypeSafe API key, and costs depend on Jev: the README quotes roughly $0.02 to $0.03 per hour of video screened. It says embeddings suit repeated queries over a large library better, and the two can be combined.
Install
Run in a terminal
claude mcp add jevreel --env TYPESAFE_API_KEY='<TYPESAFE_API_KEY>' -- uvx jevreelAdd to claude_desktop_config.json
{
"mcpServers": {
"jevreel": {
"command": "uvx",
"args": [
"jevreel"
],
"env": {
"TYPESAFE_API_KEY": "<TYPESAFE_API_KEY>"
}
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"jevreel": {
"command": "uvx",
"args": [
"jevreel"
],
"env": {
"TYPESAFE_API_KEY": "<TYPESAFE_API_KEY>"
}
}
}
}Add to .vscode/mcp.json
{
"servers": {
"jevreel": {
"type": "stdio",
"command": "uvx",
"args": [
"jevreel"
],
"env": {
"TYPESAFE_API_KEY": "<TYPESAFE_API_KEY>"
}
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"jevreel": {
"command": "uvx",
"args": [
"jevreel"
],
"env": {
"TYPESAFE_API_KEY": "<TYPESAFE_API_KEY>"
}
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"jevreel": {
"command": "uvx",
"args": [
"jevreel"
],
"env": {
"TYPESAFE_API_KEY": "<TYPESAFE_API_KEY>"
}
}
}
}Using several servers? Build one config file with the MCP config generator.
Configuration
TYPESAFE_API_KEYTypeSafe API key for Jev, from https://console.typesafe.aiRequired
What it does
- Contact-sheet screening
- Checks every second of video using 60-frame labelled sheets.
- Full-frame verification
- Rechecks top windows half a second apart.
- Typed questions
- Supports yes/no, choice and score questions per minute.
- Custom lenses
- Adds your own questions to every minute of the index.
- Photo search
- Finds a person or object from a reference image.
- Dialogue search
- Uses captions or local speech-to-text.
Good for
- Find the moment someone does a specific thing in a video
- Search what is said in a recording
- Read on-screen text such as a headline
- List every scene a character appears in
Pricing
Open source
Requires a paid TypeSafe API key; about $0.02 to $0.03 per hour of video per the README
jevreel FAQ
Does it use embeddings?
No, it uses a decision model and stores no vectors.
What do I need installed?
Python 3.11 or newer, FFmpeg and a TypeSafe API key.
What does it cost?
The README quotes about $0.02 to $0.03 per hour of video screened, paid to TypeSafe for Jev.
When are embeddings better?
For repeated queries over a large library, since each lookup is then free.
Is jevreel free?
Yes. jevreel is open source, so you can use and self-host it for free.
How do I add jevreel to Claude Code?
Run this command in your terminal: claude mcp add jevreel --env TYPESAFE_API_KEY='<TYPESAFE_API_KEY>' -- uvx jevreel
What are the best alternatives to jevreel?
The closest alternatives to jevreel are HyperFrames by HeyGen, headroom, omniget, next-ai-draw-io and mcp-for-blender.
Where this comes from
- Description: written from the project's README.
- Install steps: from the official MCP Registry entry.



