# Neuroencoder EEG embeddings

> Neuroencoder EEG embeddings is an MCP server for knowledge and memory. Turn EEG recordings into 768-dimensional embeddings and run sleep staging, from an agent. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Free trial
- **Platforms:** API, CLI
- **Website:** https://docs.neuroencoder.com/agents
- **MCP Registry name:** com.neuroencoder/eeg-embeddings
- **Categories:** knowledge and memory, health and science, AI models and LLM tooling
- **Listing updated:** 2026-11-16

## About

Neuroencoder offers a hosted EEG embeddings API, and this MCP server wraps it for agents. It accepts EDF, BDF, FIF or NPY files in any montage and sampling rate and returns one 768-value vector per window, by default 30 seconds. Jobs run asynchronously: an agent requests an upload URL, queues an embedding or sleep-staging job, polls for status and fetches a result link.

Sleep staging works on whole-night polysomnography recordings of at least two hours with an EEG channel at 100 Hz or more, and the docs mark it as research use only, not a medical device. Billing counts 30 seconds of EEG per window and refunds failed jobs.

Getting started needs no human: samples are downloadable without a key, and a single signup request returns a provisional agent key good for 2 hours of EEG over 7 days. There is also a pip-installed CLI and a local MCP server that can inspect a file for free, embed it, find similar windows and plot a PCA.

## Features

- **Any EEG format:** EDF, BDF, FIF or NPY with any montage and sampling rate.
- **Window embeddings:** One 768-d vector per window, 30 seconds by default.
- **Sleep staging:** Whole-night recordings produce a hypnogram and summary for research use.
- **Agent signup:** One request returns a provisional key without an email.
- **Free samples:** Precomputed samples are downloadable without a key.
- **Local companion:** A local server inspects files free and finds similar windows.

## Good for

- Embed overnight recordings for similarity search
- Check how many windows a file will bill before uploading
- Stage sleep in a polysomnography night for research
- Compare embeddings with a reference sample

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http eeg-embeddings https://api.neuroencoder.com/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://api.neuroencoder.com/mcp
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "eeg-embeddings": {
      "url": "https://api.neuroencoder.com/mcp"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "eeg-embeddings": {
      "type": "http",
      "url": "https://api.neuroencoder.com/mcp"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "eeg-embeddings": {
      "serverUrl": "https://api.neuroencoder.com/mcp"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "eeg-embeddings": {
      "type": "http",
      "url": "https://api.neuroencoder.com/mcp"
    }
  }
}
```

## Tools

- `account`: This API key: tier, limits and usage this month.
- `list_samples`: Precomputed real outputs (no upload needed): ids, titles, sizes, download URLs.
- `get_sample`: One sample: README, licence, channels_used, and its embeddings in the POST /v1/embeddings response format (first windows only for long ones).
- `create_upload`: Start an upload of a local .edf, .bdf or .npy file (at most 2 GB). Returns id and a presigned upload_url: PUT the file there within 10 minutes with exactly upload_headers (e.g. curl -X PUT --upload-file night.edf -H 'Content-Length: N' '<upload_url>', no Authorization header), then call start_embedding_job with the id. Only the extension of filename is used.
- `start_embedding_job`: Queue an embeddings job on an uploaded file. window_seconds: 2 to 30 (default 30); stride_seconds: default window_seconds. picks: EDF/BDF channel labels to keep; channel_names: 10-20 labels overriding the file's; sfreq: required for .npy. Bills 30 s of EEG per window at any window_seconds (refunded on failure). Poll get_job.
- `start_sleep_staging_job`: Queue sleep staging of an uploaded whole-night PSG (.edf/.bdf; >= 1 EEG channel with a 10-20 label at >= 100 Hz; EOG used if present). eeg_channel/eog_channel: channels for the YASA features; picks, channel_names as for embeddings. Counts 1 night of the weekly quota (refunded on failure). Poll get_job, then get_result_url. Research use only, not a medical device.
- `get_job`: A job's status (queued, running, succeeded, failed, cancelled, expired), usage, error or result. Poll every 5 s while queued or running.
- `get_result_url`: Fresh presigned GET URLs (10 minutes) of a succeeded job's result. Embeddings: .npy float32 [windows, 768], window i starts at i * t_start_step s. Sleep staging: url = hypnogram JSON (stage, probabilities, confidence per 30 s epoch), csv_url = CSV, plus the summary.

## FAQ

### Do I need a key to look around?

No. Samples can be fetched without one, and an agent key comes from one signup request.

### How is usage billed?

30 seconds of EEG per window; failed jobs are refunded, and sleep staging counts one night of a weekly quota.

### What files are accepted?

EDF, BDF or NPY up to 2 GB through uploads; FIF is listed in the registry description.

### Is sleep staging medical?

No. The docs say it is for research use only and not a medical device.

### Is Neuroencoder EEG embeddings free?

Neuroencoder EEG embeddings is a paid product with a free trial.

### How do I add Neuroencoder EEG embeddings to Claude Code?

Run this command in your terminal: claude mcp add --transport http eeg-embeddings https://api.neuroencoder.com/mcp

### What are the best alternatives to Neuroencoder EEG embeddings?

The closest alternatives to Neuroencoder EEG embeddings are Codebase Memory, XERJ, Claude Code Ultimate Guide, Tooluniverse and Nautilus Compass Memory.

## Alternatives

- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Tooluniverse](https://ai-powered.com/mcp/tooluniverse): Tooluniverse is an open-source MCP server for health and science.
- [Nautilus Compass Memory](https://ai-powered.com/mcp/nautilus-compass-memory): Nautilus Compass Memory is an open-source MCP server for knowledge and memory.
- [Oya Browser](https://ai-powered.com/mcp/oya-browser): Oya Browser is a freemium MCP server for browser automation.
- [note1](https://ai-powered.com/mcp/note1): note1 is an MCP server for productivity.
- [Later Stages](https://ai-powered.com/mcp/later-stages): Later Stages is a free MCP server for health and science.

---
Source: https://ai-powered.com/mcp/neuroencoder-eeg-embeddings
