Humalike
APIs that give chatbots and AI agents a sense of social timing — when to wait, when to interrupt, when to remember what you said last week — instead of the usual stiff turn-by-turn replies.
🔗 Visit Humalike
Description
Talk to most AI chatbots for more than a minute and something feels off: they answer instantly no matter what, forget the joke you made two messages ago, and never quite read the room. Humalike sells the missing piece — a set of APIs that plug into any AI agent to add the small social behaviors humans do without thinking, so the agent feels less like a form and more like it's actually paying attention.
Humalike is behavioral infrastructure for AI agents: composable APIs for turn-taking, theory of mind, social memory, norm adaptation and persona consistency, usable in both 1:1 chats and group settings. It's model- and stack-agnostic, meaning it layers on top of whatever LLM a team already uses rather than replacing it. The company also ships a Hermes plugin (117 GitHub stars) as a drop-in social-intelligence layer for that agent framework specifically. It's SOC 2 Type II certified and GDPR compliant, signaling it's built to be embedded in production products, not just demos.
💬 Our review
The short version: if your AI agent's biggest complaint from users is "it feels robotic," Humalike is a narrowly-targeted fix — not a chatbot builder, a behavior layer for chatbots you already have.
The pitch is credible because it's specific: turn-taking and social memory are real, well-studied gaps in most agent frameworks, and treating them as a pluggable API rather than a prompt-engineering hack is the right instinct — prompting an LLM to "remember context and pace itself" is brittle, a dedicated behavioral layer is more consistent. The pay-as-you-go pricing ($0.002/interaction) is cheap enough to test without commitment. The catch is that this is infrastructure for teams already building agent products, not a tool an individual uses directly — there's no visible product to click around in, just docs and an integration. It's also unproven at scale: no disclosed launch date, no named customers beyond the Hermes plugin ecosystem, so the theory-of-mind and social-memory claims are Humalike's own framing until third parties validate them.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Gratuit avec 10 000 interactions/mois (20$ de crédits) ; Pro à 0,002$/interaction ; Enterprise sur devis
Pros
API composables ciblées (turn-taking, mémoire sociale, théorie de l'esprit)
Agnostique au modèle et à la stack existante
Tarification pay-as-you-go accessible pour tester
SOC 2 Type II + GDPR
Cons
Pas de produit visible à essayer sans intégrer du code
Pas de date de lancement ni de clients nommés au-delà du plugin Hermes
Bénéfice difficile à mesurer objectivement (qualité "sociale" subjective)