nanobot + Honcho: Long-Term Memory for Your AI Assistant

A fork of HKUDS/nanobot with Honcho AI-native memory built in.

---

nanobot is an ultra-lightweight personal AI assistant (~4,000 lines of core code). This fork adds Honcho for persistent, cross-session user memory — your bot remembers who you are, what you care about, and what you've talked about, without you having to repeat yourself.

What Honcho Adds

• Cross-session memory — context carries over between conversations automatically
• User modeling — builds a representation of the user over time (preferences, history, patterns)
• Prefetch context — relevant user context is injected into the system prompt before each LLM call
• Session rotation — /clear or /new starts a fresh conversation while preserving long-term memory
• Automatic migration — existing local sessions and memory files migrate to Honcho on first use

Honcho is enabled by default on this fork. The bot works without it (graceful degradation), but memory won't persist across sessions.

Quick Start

git clone https://github.com/plastic-labs/nanobot-honcho.git
cd nanobot-honcho
pip install -e ".[honcho]"
nanobot onboard

Configure

Add your LLM provider key and Honcho key to ~/.nanobot/config.json:

{
  "providers": {
    "openrouter": {
      "apiKey": "sk-or-v1-..."
    }
  }
}

Set up Honcho (free key from app.honcho.dev):

nanobot honcho enable --api-key YOUR_HONCHO_KEY

Chat

nanobot agent                    # interactive mode
nanobot agent -m "Hello!"       # single message
nanobot gateway                  # start gateway (Telegram, Discord, etc.)

Slash Commands

Command	Description
/new	Start a new conversation (consolidates memory)
/clear	Clear session and start fresh (rotates Honcho session)
/help	Show available commands

Configuration

Config file: ~/.nanobot/config.json

Honcho

Field	Default	Description
honcho.enabled	true	Enable Honcho memory integration
honcho.workspaceId	"nanobot"	Honcho workspace identifier
honcho.prefetch	true	Inject user context into system prompt
honcho.environment	"production"	Honcho environment

Honcho CLI

nanobot honcho enable --api-key KEY   # Enable + save API key
nanobot honcho enable --migrate       # Migrate local sessions to Honcho
nanobot honcho disable                # Revert to local file-based sessions
nanobot status                        # Check Honcho connection status

Providers

This fork supports all upstream providers. See the upstream README for the full list:

OpenRouter, Anthropic, OpenAI, DeepSeek, Groq, Gemini, MiniMax, AiHubMix, DashScope, Moonshot, Zhipu, vLLM, OpenAI Codex (OAuth), and any OpenAI-compatible endpoint.

Channels

All upstream channels are supported. See upstream docs for setup:

Telegram, Discord, WhatsApp, Feishu, Mochat, DingTalk, Slack, Email, QQ.

MCP

nanobot supports MCP for external tool servers. Config format is compatible with Claude Desktop / Cursor. See upstream docs.

How It Works

User message
    |
    v
[Honcho prefetch] -- retrieve user context from memory
    |
    v
[System prompt + user context + conversation history]
    |
    v
[LLM generates response]
    |
    v
[Honcho sync] -- save exchange to persistent memory
    |
    v
Response to user

On first message per session, local JSONL history and MEMORY.md/HISTORY.md files are automatically migrated to Honcho. After migration, Honcho is the single source of truth for conversation persistence.

Upstream

This fork tracks HKUDS/nanobot and merges upstream regularly. The Honcho integration is maintained as an additive layer — all upstream features (MCP, providers, channels, skills, cron, heartbeat) work as documented.

To contribute Honcho-specific changes, open a PR against this repo. For general nanobot improvements, consider contributing upstream.

License

MIT — same as upstream.