Ship agents people actually want in the room.
Behavioral infrastructure for humanlike AI agents.
AI still feels like AI.
Smart isn't enough.
Today's models are capable enough. Smart enough. Fast enough. But we still feel they're not us. So AI gets adopted where the experience of using it doesn't matter — code, search, automation. And stalls everywhere it does.
Companions, NPCs, classrooms, voice agents — none have broken through. Not for lack of features. The interaction still feels off. That's not a feature problem. It's a behavior problem.
Humalike is building the behavioral infrastructure for humanlike AI agents. The social skills your agents have been missing. APIs, models, benchmarks.
Behavioral APIs
Coming soonBehavioral APIs you can build with.
Seven behavioral components — model, use-case, and stack agnostic — designed to compose into any agent that needs to feel human. Built for 1:1 and groups: same APIs whether it's one person, two, or ten.
Flagship
Turn-taking
Our flagship API — the rhythm of real conversation. Knowing when to speak, when to wait, when to interrupt — with norms, theory of mind, and memory composed underneath.
Norms
Picking up inside jokes, hidden rules, and tone from any group. Insider, not stranger.
Persona
Personas grounded in real community data — not generic backstories.
Theory of Mind
Tracking what each person believes, wants, and feels as the conversation unfolds.
Social Observability
Seeing how your agent lands — who likes it, who's bored, who's annoyed.
Social Memory
Remembering norms, opinions, and behavior of every person across every conversation.
Social Signals
Reading the cues most agents miss — typing pauses, edits, deleted reactions.
At Humalike, we build what makes AI fit in among humans — not just answer them.
Use cases
Where the behavior layer lands.
The places AI has to feel human — gaming squads, classrooms, hardware companions, online communities. Each one is a scenario where features alone fall short and behavior is the unlock.
AI Gaming Characters
NPCs, teammates, opponents — characters that talk, react, and remember who you are across sessions.
ViewAI Coworkers
Agents that own tasks, join your channels, and brainstorm alongside you — like a real coworker, not a prompt box.
ViewAI Therapists
Mental-health support agents that catch distress signals, hold patient context, and know when to escalate to a human.
ViewAI Companions
Pendants, speakers, glasses — hardware companions that remember context, sense mood, and know when to speak.
ViewEd-Tech
Agents that study alongside learners, debate ideas, and make courses feel like something worth showing up to.
ViewAI Closers
Sales agents that catch hesitation, remember each buyer, and pitch in your team's voice.
ViewAI Streamers & Communities
Agents that socially react, joke, and hype — keeping communities alive and chat moving.
ViewBuild with our APIs — or our team.
Designing a Deceptively Human AI Facilitator for Group Chats.
Most conversational agents are built for one-on-one, turn-based exchanges — not natural, asynchronous group chats. Humalike is an LLM-based facilitator that joins multi-party conversations using humanlike strategies and timing.
Recognizing Social Norms and Sanctioning in Group Chats.
Ask AI about Humalike.
Pick your favorite AI and ask what it thinks about Humalike.
Their words, not ours.