MEMORY THATREASONS
MEMORY THAT REASONS
Continual learning for stateful agents. Better context. Fewer tokens.
npx skills add plastic-labs/honcho
Run in your terminal to install the Honcho skill. Learn more
Beyond Store & Retrieve
Memory systems store facts. Users and agents are more than facts. And facts waste tokens. Delivering the context that matters requires reasoning.
Honcho doesn't just store data. It continually learns. Every message triggers comprehensive reasoning that saves tokens downstream:
Our reasoning models achieve SOTA performance at lower cost and latency than frontier models. Neuromancer powers learning beyond explicit facts, it reasons toward conclusions that follow, patterns across interactions, and hypotheses to test against new data.
The result: token savings and richer user context.
Statefulness, Solved
Memory is solved. Retrieval is unlimited. Stateful agents with a single method call.
How it works
Write messages to Honcho and two things happen:
- Messages are stored and indexed
- Neuromancer reasons and learns automatically
Just call get_context() for effortless state.
get_context() returns instantly curated reasoning plus conversation history. Everything an agent needs to maintain continuity and efficient enough to use at every turn.
Granular control. Shape retrieval with search, token budgets, summaries, and peer scoping.
Explore the docsAny stack. Any scale. Honcho just works.
Advanced Reasoning On-Demand
Call .chat() for specific queries. More reasoning resources when it matters.
Minimal
Single semantic search. One lookup.
Low
Conclusions with surrounding context. Default tier.
Medium
Multiple searches. Directed synthesis.
High
Multi-pass analysis. Patterns over time.
Max
Research-grade. Exhaustive full history search with quantitative methods.
Transcend Memory
Honcho is more than memory. It enables interaction patterns that weren't possible before.
Peers
Other memory systems restrict use to the user-assistant paradigm. Honcho is more dynamic. It can learn about any entity, modeling users, agents, NPCs, groups, and their relationships.
Peers can be freely added or removed from Sessions, and their perspectives scoped global or surgically. Designed for ultimate flexibility.
Dreaming
Honcho also reasons in the background. Dreaming extends continual learning.
Asynchronous reasoning runs automatically, continuously optimizing Honcho's understanding of each Peer without impacting runtime performance.
Standard dreaming is included with every workspace. Enhanced dreaming options are available for deeper reasoning use cases.
Standard dreaming is included with every workspace. Enhanced dreaming options are available for deeper reasoning use cases.
Stateful Production
Honcho's primitives support diverse agents. Vibecode to scale and vertical-agnostic.
Enable what matters: emotional arcs, shared history, implicit meaning, preference alignment. No more "dementia." Build lasting connections.
Learn team conventions and architecture philosophy. Anticipate intent. Code that matches your style, not generic best practices.
NPCs form opinions, track relationships, build identites, and reshape narratives without $0.50/interaction costs breaking the budget.
Track misconceptions across sessions. Predict obstacles before they stall progress. Tutoring that compounds over semesters.
Customer history persists across sessions, channels, agents. No more "can you start from the beginning?"
Workflow context persists across tools and time. Needs aren't just met, they're anticipated.
Transparent. Predictable.
Use the 10K tokens you need, not the 100K you don't.
HONCHO MEMORY
HONCHO REASONING
Startups (<$5M raised)
$1,000 in credits. 12 months subsidized pricing. Integration support. Grow with Honcho.
Apply NowEnterprise
Custom plans. Forward-deployed engineers. Dedicated integration and maintenance support.
Contact Founders