Honcho

What is Honcho

Honcho is an AI-native memory platform for stateful agents and continual context. It provides memory that reasons, enabling continual learning for stateful agents with better context and fewer tokens. Honcho is developed by Plastic Labs and is designed to go beyond simple store-and-retrieve memory systems by continually learning from every message.

How to use Honcho

1. Start Building: Access the Honcho app to begin building with the platform.
2. Vibecoding Setup: Follow the vibecoding setup guide for a streamlined development experience.
3. Install the Honcho CLI: Use the Honcho CLI for command-line interactions.
4. Integrate with Your Tools: Install plugins for various platforms:
   • Agent Skill: Run npx skills add plastic-labs/honcho in your terminal.
   • Claude-Honcho: In Claude Code, run /plugin marketplace add plastic-labs/claude-honcho and /plugin install honcho@honcho.
   • Codex-Honcho: Run npm install -g @honcho-ai/codex-honcho and codex-honcho install, then restart Codex.
   • OpenClaw: Run openclaw plugins install @honcho-ai/openclaw-honcho and openclaw gateway restart.
   • Hermes Agent: Install Hermes via curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash and run hermes memory setup honcho.
   • DeepSeek Harness: Run dsh plugin --profile <name> add @honcho-ai/dsh-honcho.

Features of Honcho

• AI-Native Memory: Memory that reasons, not just stores and retrieves.
• Continual Learning: Every message triggers comprehensive reasoning that saves tokens downstream.
• Stateful Agents: Supports continual learning for stateful agents.
• Better Context, Fewer Tokens: Optimizes context delivery to reduce token usage.
• Custom Models: Includes custom models for memory reasoning.
• Integrations: Works with Claude Code, OpenAI Codex, OpenClaw (WhatsApp, Telegram, Discord, Slack), Hermes Agent, and DeepSeek Harness.
• CLI: Provides a command-line interface for management.
• Agent Skill: Installable skill for developers integrating Honcho.

Use Cases of Honcho

• Persistent Memory for Claude Code: Adds long-term memory to Claude Code sessions.
• Persistent Memory for OpenAI Codex: Enhances Codex with memory capabilities.
• Memory for Messaging Platforms: Integrates with WhatsApp, Telegram, Discord, and Slack via OpenClaw.
• Autonomous Agents: Powers autonomous agents like Hermes Agent with built-in Honcho memory.
• DeepSeek Harness: Provides persistent memory for DeepSeek Harness.
• Stateful Agents: Enables continual learning and context retention for AI agents.

FAQ

What is Honcho?
Honcho is an AI-native memory platform for stateful agents and continual context.

How does Honcho differ from traditional memory systems?
Honcho doesn't just store data; it continually learns. Every message triggers comprehensive reasoning that saves tokens downstream.

What platforms does Honcho integrate with?
Honcho integrates with Claude Code, OpenAI Codex, OpenClaw (WhatsApp, Telegram, Discord, Slack), Hermes Agent, and DeepSeek Harness.

Is Honcho state-of-the-art?
Yes, Honcho is state-of-the-art on LongMem, LoCoMo, and BEAM benchmarks, and is Pareto dominant in accuracy, cost, speed, and tokens.