Nefesh — Real-Time Human State Awareness for AI
by nefesh.ai · remote · no sign-in to list tools · v4.0.0 · published by ai.nefesh
Fuses biometric signals into a stress score (0-100) for AI adaptation. MCP + A2A native.
Deletes · 1Final · 1Changes · 1
7 tools: 1 can delete or cancel things, 1 do something that can't be undone, 1 change things and 4 only read.
Scanned Oct 21, 2026
- Deletes or cancels1Claude Code: deny: see these tools
- Can't be undone1Claude Code: ask: see these tools
- Changes things1Claude Code: ask: see these tools
- Only reads4Claude Code: allow: see these tools
No check matched anything in the tool definitions. No finding is not a safety guarantee. How the checks work.
claude mcp add --transport http human-state https://mcp.nefesh.ai/mcpTools
Open a tool for its reasons and full description. Descriptions are the publisher's own words, shown as they are.
7 tools
Deletes or cancels 1Claude Code: deny
delete_subjectPermanently delete all sessions and trigger memory for a subject.No findings
Why it is in this group
- The name "delete_subject" says "delete": a word for removing or ending something.
- The description says "Permanently delete".
- The server marks it destructive (destructiveHint: true).
Description, as the server wrote it Permanently delete all sessions and trigger memory for a subject. Irreversible. Only call when the user explicitly requests this deletion.
Annotations the server declares
- destructiveHint: true
- idempotentHint: true
Can't be undone 1Claude Code: ask
ingestSend biometric signals from any sensor, get unified state back.No findings
Why it is in this group
- The description's first verb is "send": a word for sending, and what is sent cannot be taken back.
Description, as the server wrote it Send biometric signals from any sensor, get unified state back. Required: session_id + timestamp (ISO 8601) + at least one signal. Send whatever you have — the API fuses all signals into one state. Common signals (highest impact): - heart_rate (bpm, 30-220) + rmssd (ms) — cardiovascular - tone: calm | tense | anxious | hostile — vocal - sentiment: -1.0 to 1.0 — textual - expression: relaxed | neutral | tense — visual For trigger memory (cross-session psychological tracking): - Include subject_id (consistent per user, hashed) Returns same fields as get_human_state plus signals_received list and topics_detected (if conversation text was included). source_device is optional but improves confidence scoring. Not a medical device.
Changes things 1Claude Code: ask
request_api_keyStart free API key authorization; no existing API key needed.No findings
Why it is in this group
- The name "request_api_key" says "request": a word for a change.
Description, as the server wrote it Start free API key authorization; no existing API key needed. Returns a verification_url. Ask the user to open that URL in their browser, enter their own email, then open the verification email and confirm access for their agent. This tool does not accept email addresses and does not send email. Do not choose, invent, or submit an email on the user's behalf. Poll check_api_key_status with the returned request_id at the supplied interval. Stop if the request expires. Never create new requests in a loop.
Only reads 4Claude Code: allow
check_api_key_statusCheck a pending API key authorization using its request_id.No findings
Why it is in this group
- The name "check_api_key_status" says "status": a word for reading.
Description, as the server wrote it Check a pending API key authorization using its request_id. Use request_id exactly as returned by request_api_key, never an email. Poll every 10 seconds. awaiting_user means the user must open the URL; pending means email verification is still required. ready returns the API key once only. Store it and include it in all subsequent HTTP calls. Stop on expired; ask the user before starting another authorization.
get_human_stateGet current unified human state for a session.No findings
Why it is in this group
- The name "get_human_state" says "get": a word for reading.
Description, as the server wrote it Get current unified human state for a session. Call this before generating important responses. Returns: - state: calm | relaxed | focused | stressed | acute_stress - stress_score: 0-100 (lower = calmer) - confidence: 0.0-1.0 (based on signal quality and device type) - suggested_action: maintain_engagement | simplify_and_focus | de-escalate_and_shorten | pause_and_ground - action_reason: human-readable explanation of why this action was suggested - adaptation_effectiveness (on 2nd+ call): shows whether your previous suggested_action actually reduced stress — contains previous_action, stress_delta, and effective boolean. Use this to self-improve. Use suggested_action to adapt your response: calm/relaxed = full complexity, focused = shorter and structured, stressed = max 2 sentences, acute_stress = one grounding sentence only. Requires a prior ingest call to have data. Not a medical device.
get_session_historyGet state history for a session over time.No findings
Why it is in this group
- The name "get_session_history" says "get": a word for reading.
Description, as the server wrote it Get state history for a session over time. Returns timestamped datapoints with stress_score, state, and heart_rate for each observation. Includes an overall trend: rising | falling | stable. Use minutes parameter to control the lookback window (default: 5, max: 60). Useful for detecting stress patterns during a conversation. Not a medical device.
get_trigger_memoryRetrieve psychological trigger profile for a subject.No findings
Why it is in this group
- The name "get_trigger_memory" says "get": a word for reading.
Description, as the server wrote it Retrieve psychological trigger profile for a subject. Returns which conversation topics consistently cause stress (active triggers) and which have been resolved over time. - active triggers: topics where stress was elevated across multiple sessions. Tread carefully. - resolved triggers: topics where stress has decreased. Safe to explore deeper. Each trigger includes observation_count, avg_score, peak_score, and last_seen. Requires prior ingest calls with the same subject_id. Not a medical device.
This is not a safety rating, and no finding is not a safety guarantee. The groups and findings come from fixed checks of the tool definitions the server returned on the date shown, not from its code or what it does when called. How the checks work.
Install
Pinned to version 4.0.0. Values in angle brackets are yours to fill in.
claude mcp add --transport http human-state https://mcp.nefesh.ai/mcpRun in a terminal.
Claude Code permissions
Allow the reads, ask before a change, deny the deletes. Paste into .claude/settings.json.
{
"permissions": {
"allow": [
"mcp__human-state__get_human_state",
"mcp__human-state__get_trigger_memory",
"mcp__human-state__get_session_history",
"mcp__human-state__check_api_key_status"
],
"ask": [
"mcp__human-state__ingest",
"mcp__human-state__request_api_key"
],
"deny": [
"mcp__human-state__delete_subject"
]
}
}Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. It assumes the server is installed as human-state. Review it before you use it.