Back to Gap Map Product / UX / Agent memory & context stores

Honcho

Plastic Labs
open source / Overall score: 2.8

Honcho is a memory service for stateful agents that stores the messages exchanged between users and agents, reasons over them in the background to derive conclusions about each participant, and serves representations, session summaries and search results back as context. It runs as a FastAPI server, self-hosted or managed, with Python and TypeScript SDKs and plugins for coding agents. Plastic Labs builds it.

Openness

5 high confidence
5.0
license
AGPL-3.0 (the Honcho server)+Apache-2.0 (the Python and TypeScript SDKs)+MIT (honcho-cli)
source
public
core features withheld
no — the 1,329-path tree at 4b9e27b holds the server, deriver and dreamer workers, SDKs and CLI with no ee/, enterprise/, commercial/ or proprietary/ path, and the managed service is metered on stored tokens and queries
commercial
Honcho managed service at api.honcho.dev

Honcho's server, including the background workers that derive and consolidate conclusions about users and agents, is published in full and self-hosts from the repository. Plastic Labs sells a managed service metered on stored tokens and reasoning queries, not features held back from the code.

Adoption

2 low confidence
2.0

GitHub stars are the only signal, because the Honcho server ships as source and a Docker Compose setup rather than through a package registry. The honcho-ai and @honcho-ai/sdk packages are clients that reach either the managed service or a self-hosted server, so their downloads count client installs rather than Honcho deployments and are left out. A star is not a deployment.

Capability

4 medium confidence
4.0

Honcho extracts conclusions about each user or agent from raw messages, revisits them in periodic dream cycles that replace outdated facts, and serves them as peer cards, representations and summaries. Nothing in it retires memories because of their age.

Verified 2026-11-20, except some scores, still awaiting re-confirmation