Background deriver models a person's beliefs/preferences/contradictions; a 'peer' can be a human, an agent, or an idea.
- Storage
- Cloud, or FastAPI + Postgres/pgvector + Redis + your LLM
- Retrieval
- Queries a derived user-representation rather than raw message recall.
- Self-host
- Self-host: heavy
- License
- AGPLv3
- Pricing
- $2 / 1M tokens ingested; $100 credits · Freemium
- GitHub stars
- 5,608
- Last release
- —
- Last commit
- 2026-06-26
- First catalogued
- 2026-06-28
Strengths
- Genuine user-modeling, not fact recall
- Multi-agent / perspectival via peers
- Cheap to ingest
Watch out
- Reasons in background → bills an LLM even self-hosted; AGPL; somewhat opaque
Best for
- Personalization that models a user's beliefs/preferences over time (theory-of-mind)
- Multi-party modeling — what one peer (human or agent) knows about another
Benchmark results
| Benchmark | Value | Backbone | Trust | Source |
|---|---|---|---|---|
| longmemeval | 90.4 accuracy | — | Self-reported | Honcho (Plastic Labs) ↗ |
| locomo | 89.9 accuracy | — | Self-reported | Honcho (Plastic Labs) ↗ |
Sources
- https://honcho.dev/ (vendor)
- https://github.com/plastic-labs/honcho (vendor)
- Honcho review — theory-of-mind, peer-centric (human/agent/idea) personalization (third-party)
Last verified 2026-06-28 · updated by refresh-framework-cards