v0.2.0 · Welcome
Every AI should know who it's talking to.
Humane AI is the behavioral operating system for longitudinal AI. The fastest path to value is one line — observe() returns user-state and risk with no onboarding. These docs cover everything from that five-minute first signal to the concepts behind the engine, vertical-template tutorials, and the complete API reference with an interactive explorer.
Start here
pip install humane-ai · one-line wrap for 20+ providers · async + sync clients.
npm i @humane-ai/sdk · works in Node, Deno, Bun, and the browser.
Raw HTTP if you want to integrate from any other language.
observe() call, and read r.user_state.mood / r.risk.level. No onboarding, no model config, under 90 seconds.Core concepts
No-LLM behavioral monitoring. One line, no onboarding, instant user-state + risk. Never blocks your reply.
An operator steps into a live AI conversation — the bot pauses its AI and delivers the operator's messages, then hands back.
Semantic, per-end-user memory that persists across sessions — MemPalace.
Pre-LLM classification with explainable matched-phrase audit.
The engine reaches out when mood, energy, or trust crosses thresholds.
Deterministic variant routing with two-proportion z-test analytics.
Behavior-as-code YAML rules your compliance officer can author.
Statistical z-score monitoring on sentiment, engagement, latency, and errors with webhook + email alerts.
Tutorials
Ten one-click industry bundles — Therapy, Wellness, Elder Care, Customer Support, Financial Coach, and more.
Longitudinal mental-health support with hard-wired crisis escalation.
Messaging Channels
Connect a Telegram bot via @BotFather. Verified per-channel with the X-Telegram-Bot-Api-Secret-Token header.
Wire a WhatsApp Cloud API number. HMAC-SHA256 verification over the raw body keeps the endpoint safe.
Reference
Interactive OpenAPI — try calls directly from the browser.
Flat endpoint-by-endpoint list with parameters and examples.
Every event the engine emits, with field-by-field payload documentation.
One more thing
Read the philosophy, too.
Humane was designed around one line: Every AI should know who it's talking to.The concepts section explains what that means — and why it's not something you can bolt on later.