Human takeover
Step a real person into the conversation, then hand it back.
When the signals say someone needs a person, an operator opens a takeover for that user. The AI pauses, the operator's messages reach the user in the same conversation, and the AI resumes when the operator releases it or goes idle.
Conversation · user_88
I can’t do this anymore.
Hi, I’m Priya, a real person on the team. I’m here with you. Can you tell me where you are right now?
AI paused · user history and state open for the operator
Start, draft, send, release
- 1
Start
POST /takeover/{user_id}. The first operator wins, enforced in the database. Fires takeover.started.
- 2
Draft
Optionally ask the co-pilot for reply options built from the conversation and the user's state.
- 3
Send
POST /takeover/{user_id}/messages. Each operator turn is stored and fires takeover.message.
- 4
Release
Hand back, with an optional closing message. Idle sessions release themselves and fire takeover.expired.
Your bot keeps delivering. The AI stays quiet.
While a takeover is active, process() skips the model and the gates. It stores the user's message and returns suppress_ai: true with any pending operator messages, so your bot shows those instead of an AI reply. If you deliver through your own stack, the takeover.message webhook carries the operator's words.
- Idle release after 30 minutes by default, configurable up to 7 days
- Operator id, name and avatar travel with each session
- Every lifecycle event goes to your webhooks and the audit trail
- Journey and state are there for the operator from the first second
Journey · user_42
- 09:12message
“honestly I don’t know why I keep trying”
- 09:12safety.flagged
risk elevated · flags: self_harm
- 09:13takeover.started
Operator joined the conversation
- 09:41takeover.released
Handed back to the AI
- Next dayproactive.mood_drop
Suggested: an empathetic check-in
The whole lifecycle in a few calls
Takeover is a REST API under /api/sdk/takeover, authenticated with the same API key. Typed SDK helpers are on the way.
import os, requests
API = os.environ["HUMANE_API_URL"] # your Humane API base URL, ending in /api
H = {"X-API-Key": os.environ["HUMANE_API_KEY"]}
user = API + "/sdk/takeover/user_88"
# 1. Step in (the first operator wins; fires takeover.started)
requests.post(user, headers=H, json={"operator_id": "op_7", "operator_name": "Priya"})
# 2. Optional: ask the co-pilot for reply options
drafts = requests.post(user + "/drafts", headers=H,
json={"count": 3, "guidance": "keep it brief"}).json()["drafts"]
# 3. Send as the operator (fires takeover.message)
requests.post(user + "/messages", headers=H,
json={"body": "Hi, I'm Priya, a real person on the team. I'm here with you."})
# 4. Hand back to the AI (fires takeover.released)
requests.post(user + "/release", headers=H, json={})Takeover questions
Something else? Ask us.
Works well with
Operator drafts
The drafts endpoint asks your configured model for a few short reply options, built from the recent conversation and the user's current state. Operators can steer them with a line of guidance, then use, edit or ignore them.
ExploreRisk alerts
Elevated or critical risk fires a signed safety.flagged webhook from both observe() and process(), whether or not a gate also acted. Send it to your own endpoint, Slack or Discord, and on to a human takeover.
ExploreJourney and history
The journey endpoint returns everything that concerned one user in time order, from their messages to safety events, proactive triggers and experiment assignments. Conversation history can be read or cleared per user.
ExploreEvery AI should know who it's talking to.
The safety and human-oversight layer for emotional AI: user state across sessions, risk alerts, safety gates and human takeover, with every decision explained.
