Honcho

Honcho — AI-Native Memory Platform

Honcho provides persistent memory, user modeling, and context management for LLM applications. It learns about users over time through conversations and builds rich representations.

Overview

• Workspaces: Top-level containers isolating data between applications
• Peers: Represent users/entities — Honcho builds profiles ("representations") from conversations
• Sessions: Conversation threads containing messages between peers
• Conclusions: Derived insights Honcho extracts from conversations
• Dreaming: Autonomous background process that consolidates and improves memory
• Context API: Retrieve formatted context (summaries + recent messages) for LLM injection

Quick Start

from honcho import Honcho

honcho = Honcho()
workspace = honcho.workspaces.get_or_create("my-app")
peer = honcho.workspaces.peers.get_or_create(workspace.id, "user-123")
session = honcho.workspaces.sessions.get_or_create(workspace.id, session_id="conv-1", peers=["user-123"])

# Store messages
honcho.workspaces.sessions.messages.create(workspace.id, session.id, messages=[
    {"role": "user", "content": "Hello!"},
    {"role": "assistant", "content": "Hi there!"}
])

# Get context for next LLM call
context = honcho.workspaces.sessions.get_context(workspace.id, session.id)

Core Concepts

• Representations: Rich user profiles built from conversation patterns — not just facts but behavioral understanding
• Dreaming: Background process that reviews conversations, extracts conclusions, and refines representations
• Context: Combines session summary + recent messages + peer representation into a token-budgeted payload

Documentation

Getting Started

• Overview — What Honcho is and why
• Quickstart — Setup and first API calls
• Agentic Development — MCP server, agent skills, and tools

Core Concepts

• Architecture — Data model and core concepts
• Design Patterns — Workspace/peer/session patterns
• Reasoning — How Honcho reasons over data
• Representations — Peer user models

Features

• Storing Data — Messages and conversations
• Get Context — Context retrieval for LLMs
• Chat Endpoint — Querying peer knowledge
• Advanced Features — Dreaming, filters, search, streaming

API Reference

• Introduction — Authentication and basics
• Workspaces — Workspace endpoints
• Peers — Peer endpoints
• Sessions — Session endpoints
• Messages — Message endpoints
• Conclusions — Conclusion endpoints
• Webhooks — Webhook endpoints

Guides & Integrations

• Guides Overview — All guides index
• Claude Code — Honcho + Claude Code
• MCP Server — Model Context Protocol
• LangGraph — LangGraph integration
• Discord / Telegram — Bot guides

Reference

• CLI Reference — honcho-cli terminal tool (uv tool install honcho-cli)
• SDK Reference — Python and TypeScript SDKs
• Dashboard — Honcho dashboard
• Self-Hosting — Local setup
• Configuration — Provider and infra config

Upstream Sources

• Documentation: https://docs.honcho.dev/v3/documentation
• Repository: https://github.com/plastic-labs/honcho
• Dashboard: https://app.honcho.dev
• Docs Index: https://docs.honcho.dev/llms.txt