Nefesh

Name: Nefesh
Author: nefesh-ai

@nefesh-ai

About Nefesh

Real-time human state awareness for AI agents. Fuses cardiovascular, vocal, visual, and textual signals into a unified stress score (0-100). MCP + A2A native. 7 tools, 4 A2A skills. Closed-loop adaptation feedback. Free tier available.

#health #biometrics #stress #wellness #mcp #streamable-http

Basic information

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "nefesh": {
      "url": "https://mcp.nefesh.ai/mcp",
      "headers": {
        "X-Nefesh-Key": "<YOUR_API_KEY>"
      }
    }
  }
}

Tools

Get current unified human state for a session. Returns stress level, active signals, confidence, and a default LLM behavior recommendation. Call before generating important responses. Not a medical device.

Send human signals from any sensor, get unified state back. session_id + timestamp + at least one signal required. Heart rate, glucose, EEG, voice tone, facial expression — send what you have. Not a medical device.

Get state history over time for a session. Returns timestamped state snapshots from the last N minutes. Useful for tracking stress trends. Not a medical device.

GDPR-compliant deletion of all data for a subject. Cascading delete: removes all sessions and stored signals for the given subject_id. Irreversible. Not a medical device.

Overview

What is Nefesh?

Nefesh is an MCP server that gives AI agents real-time awareness of human physiological state. It accepts sensor data (heart rate, voice, facial expression, text sentiment) and returns a unified state with a machine-readable action for the agent, including adaptation_effectiveness on subsequent calls for closed-loop feedback.

How to use Nefesh?

Add the server URL to your MCP client config (e.g., ~/.cursor/mcp.json). Use Option A: connect without an API key and ask your agent to call request_api_key — after clicking a verification email, the agent retrieves the key automatically. Option B: sign up at nefesh.ai/signup for a free key (1,000 calls/month) and include it in the config headers. All agents connect via Streamable HTTP — no local installation required.

Key features of Nefesh

Seven tools including ingest, get_human_state, and delete_subject
Self‑provisioning: agent fetches API key autonomously after email verification
Returns adaptation_effectiveness on second+ call for closed‑loop feedback
Free tier: 1,000 calls/month with no credit card
No video/audio uploads; edge processing runs client‑side
GDPR/BIPA compliant with delete_subject for cascading deletion

Use cases of Nefesh

AI coaching agents that adapt tone and timing based on user stress state
Productivity assistants that pause or reframe tasks when detecting cognitive load
Companion agents that respond to emotional signals with appropriate empathy
Autonomous interview bots that modify pacing based on candidate facial expression

FAQ from Nefesh

How does Nefesh handle data privacy?

No video or audio is uploaded — edge processing runs client‑side. No PII is stored. The delete_subject tool enables cascading deletion for GDPR/BIPA compliance. Nefesh is not a medical device.

What are the rate limits and pricing?

Free tier: 1,000 calls/month, all signal types, 10 requests per minute. Paid Solo plan: $25/month for 50,000 calls. Enterprise plan offers custom SLA and pricing.

What transport does Nefesh use?

Streamable HTTP — no local installation needed. All MCP clients connect via a URL.

How does self‑provisioning work?

The agent calls request_api_key(email) without an API key. After you click a verification link, the agent polls check_api_key_status every 10 seconds and receives the key automatically.