Stores and retrieves persistent AI conversation memory using the Honcho API.
npx skills add https://github.com/plastic-labs/honcho --skill honcho-memory
Please help me install this Agent Skill. Skill: honcho-memory Source: https://github.com/plastic-labs/honcho/tree/main/examples/zo Command: npx skills add https://github.com/plastic-labs/honcho --skill honcho-memory
AI agents forget everything between conversations, forcing users to repeat preferences, history, and context. This Skill gives agents persistent memory by saving messages to Honcho and recalling them later through natural language queries or formatted context injection.
save_memory, then in a later session calls query_memory to recall that the user is learning Rust and enjoys rock climbing, personalizing its responses without the user repeating themselves.Set your HONCHO_API_KEY environment variable, then ask the AI to save this conversation to Honcho memory and recall what it knows about you in the next session.
High-intent search queries and answers about installing and using this skill.
Use the save_memory function to store each user and assistant message in a Honcho session, then call query_memory or get_context in later sessions to recall stored information. Memory persists across conversations as long as you reuse the same user_id.
Call query_memory with a user_id and a natural language question such as "What are my hobbies?". It uses Honcho's Dialectic API to return an answer grounded in the peer's stored observations, optionally scoped to a specific session_id.
The skill requires Python 3.9 or higher, honcho-ai version 2.1.0 or above, and python-dotenv for loading environment variables. You also need a Honcho API key set as HONCHO_API_KEY, obtainable from honcho.dev.
Yes, set the HONCHO_WORKSPACE_ID environment variable to isolate memory into separate workspaces. Using unique workspace IDs per user group is recommended in production, since shared workspaces may leak context between peers during queries.
In shared workspaces, the Dialectic API falls back to workspace-level context when the queried peer has no stored memory yet. Assign unique HONCHO_WORKSPACE_ID values per user group and save messages before querying to prevent cross-peer data leakage.
plastic-labsConfigure Honcho memory by setting an API key and validating the connection.
HouseGarofaloExtract structured data from LLM responses into Pydantic models.
coco-deDesign and execute cohort analysis for user retention and engagement.
cxnaiveStore and retrieve vector embeddings with metadata filtering for RAG applications.