Honcho

Founded by Vince Trost in 2023

Store and retrieve personalized memory for AI agents across conversations

Pricing model

Paid

Rating

★ People love it

Time to value

Quick Setup (< 1 hour)

You can use Honcho to add persistent, reasoning-based memory to AI agents and chatbots. It continuously learns from every message, builds user context, and returns curated summaries and conversation history with a single API call. It supports token budgets, background reasoning (called Dreaming), multi-entity memory (users, agents, groups), and works with OpenAI, Anthropic, and custom models. Pricing scales from free memory ingestion to research-grade reasoning queries.

What Honcho does

Store and index conversation messages automaticallyRetrieve curated user context with a single API callSet token budgets to control how much context is returnedRun background reasoning to deepen user understanding asynchronouslyQuery specific user facts with on-demand reasoning at different cost tiersModel relationships between users, agents, and groups using peer scopingIntegrate persistent memory into Claude, OpenAI, or custom model agentsGenerate behavioral profiles and sentiment trend analysis from conversation historyAutomatically reasons over every incoming message to build user context without manual taggingReturns curated memory and conversation history in ~200ms with a single context() callBackground Dreaming feature runs async reasoning to deepen understanding without slowing runtimeSupports multi-entity memory: users, agents, NPCs, groups, and their relationshipsToken budgeting lets you set limits and Honcho optimizes what context gets returnedState-of-the-art benchmark scores on LoCoMo, LongMem, and BEAM evaluationsOn-demand reasoning tiers from instant $0.001 lookups to $0.50 research-grade analysisModel-agnostic: works with OpenAI, Anthropic, and custom models