StartupNew York, NY · Est. 2023
Honcho (Plastic Labs)
User modeling and psychology for agent memory.
Honcho is a user-context and memory server that models what each user knows and wants, so agents can personalize without teams hand-rolling their own user-modeling stack.
Pros & cons
Each point cites its sources and says what kind of evidence backs it. A second reviewer tried to refute every claim before it was kept.
Pros
- MetadataOpen source with about 7.8k stars, and self-hostable with Docker Compose, Postgres with pgvector and one LLM key.[1]
- DocsModels peers (users, agents, groups) and their relationships, and runs background "dreaming" outside the request path to resolve conflicts.[1]
- DocsManaged pricing is published and usage-based: $2 per million ingested tokens, unlimited context() reads, and $0.001 to $0.50 per reasoning query.[1]
- DocsStartups that have raised under $5M can get $1,000 in credits and 12 months of subsidized pricing.[1]
Cons
- MetadataAGPL-3.0 requires modifications served over a network to be open-sourced, which many companies' legal teams restrict.[1]
- DocsIts benchmark scores (89.9% LoCoMo, 90.4% LongMem S) are self-reported and come with no baseline scores, though a benchmark repo is linked.[1]
- DocsSelf-hosting needs an API server plus a separate deriver worker, and auth is off by default.[1]
- MetadataPull requests are limited to maintainers, so outside contributors can only file issues.[1]
- Best for
- Builders of personal or social agents who need detailed per-user and multi-peer modeling.
- Watch out for
- Have legal review the AGPL before you self-host it inside a commercial service.
Assessed Oct 29, 2026 · adversarially reviewedSuggest a correction
People
Co-founder & CEO
Context engineering & memory
On X
No posts from them in our X feed yet.
Neighbours in the directory
Berkeley, CA · Est. 2024
Stateful agents with memory, from the team behind MemGPT.
Letta (formerly MemGPT, out of UC Berkeley) builds an agent server with built-in memory, context management and state — letting agents persist, learn and be edited like files rather than living in a chat window.
$10M seed led by Felicis (2024)
San Francisco, CA · Est. 2024
The memory layer for AI apps and agents.
Mem0 (from the EmbedChain team) provides a managed + open-source memory layer that extracts, stores and retrieves user and entity memories across sessions — a standard building block for personalization.
San Francisco, CA · Est. 2023
Long-term memory for agents, built on temporal knowledge graphs.
Zep maintains durable, time-aware agent memory via its open-source Graphiti engine — facts, relationships and their history — so conversations and tasks can span sessions.
Connections
Relations from the Harness Hub knowledge graph, as charted in the State of Harness reports.
Similar entries · 4
- Episodic MemoryProject
- HindsightProject
- Mem0Project
- Mem0Startup