nanobot

A production-grade AI agent platform written in pure Rust.

One binary. Six channels. Fifty tools. Zero cold-start drama.

Live Demo · API Docs · Report Bug

nanobot is a self-hostable, multi-channel AI assistant that ships as a single Rust binary. It connects to 8+ LLM providers with automatic failover, exposes 50+ agentic tools, and deploys to AWS Lambda for pennies. It powers chatweb.ai and teai.io in production today.

Why nanobot?

Architecture

                         +------------------+
                         |   Your users     |
                         +--------+---------+
                                  |
            +----------+----------+----------+----------+
            |          |          |          |          |
          Web       LINE    Telegram    Discord    Slack ...
            |          |          |          |          |
            +----------+----------+----------+----------+
                                  |
                       +----------v----------+
                       |   API Gateway /     |
                       |   Reverse Proxy     |
                       +----------+----------+
                                  |
                       +----------v----------+
                       |     nanobot         |
                       |  (single binary)    |
                       |                     |
                       |  +-- Auth & Credits |
                       |  +-- Agentic Loop   |
                       |  +-- Tool Runtime   |
                       |  +-- STT / TTS      |
                       |  +-- Memory Engine  |
                       +----+------+----+----+
                            |      |    |
               +------------+   +--+    +------------+
               |                |                    |
        +------v------+  +-----v------+  +---------v---------+
        | LLM Providers|  |  DynamoDB  |  |   External APIs   |
        | (8+ w/ fail- |  | (sessions, |  | (Brave, Jina,     |
        |  over)       |  |  memory,   |  |  OpenAI TTS, ...) |
        +--------------+  |  credits)  |  +-------------------+
                          +------------+

Features

Multi-LLM with Automatic Failover

nanobot doesn't lock you into a single provider. Configure multiple API keys and it handles the rest -- round-robin load balancing, circuit breakers, and transparent failover across providers.

Tiered model selection (economy / normal / powerful) lets you balance cost and quality per request.

50+ Built-in Tools

Agentic mode executes multi-step tool chains automatically. Free users get 1 iteration; Pro users get up to 5 with parallel tool execution.

Skill Marketplace

Users can publish and install custom skills:

• Prompt skills -- inject system prompts for specialized personas or domain knowledge
• Tool skills -- expose any HTTPS endpoint as an LLM-callable tool via webhook

Skills are stored in DynamoDB and loaded at chat time. No redeploy required.

Multi-Channel

One codebase serves all channels. Conversations sync across them.

Voice-First

• STT: Web Speech API (browser-side, zero server cost)
• TTS: OpenAI tts-1 with response caching
• Auto-TTS: Voice input triggers automatic voice output
• Push-to-talk UI with visual feedback

Long-Term Memory

Two-layer auto-consolidation inspired by OpenClaw:

Session context (20 messages)
        |
        v
Daily log (auto-appended after each conversation)
        |
        v
Long-term memory (consolidated summaries)

Memory persists across channels and sessions via DynamoDB.

A/B Testing Framework

Built-in CRO experimentation:

• Deterministic variant assignment (hash(uid + testId) % N)
• Event tracking via POST /api/v1/ab/event
• Aggregated stats with 90-day TTL
• No external analytics dependency

Quick Start

Try the hosted API (no setup)

curl -X POST https://chatweb.ai/api/v1/chat \
  -H "Content-Type: application/json" \
  -d '{"message": "What can you do?", "session_id": "demo"}'

Run locally

git clone https://github.com/yukihamada/nanobot.git
cd nanobot

# Set at least one provider key
export ANTHROPIC_API_KEY=sk-ant-...
# or: export OPENAI_API_KEY=sk-...
# or: export OPENROUTER_API_KEY=sk-or-...

# Build and run the web gateway
cargo build --bin chatweb
./target/debug/chatweb gateway --http --http-port 3000
# Open http://localhost:3000

Docker

docker run -p 3000:3000 \
  -e OPENAI_API_KEY=sk-... \
  ghcr.io/yukihamada/nanobot

CLI usage

# Interactive agent mode (uses your local API keys directly)
./target/debug/chatweb agent

# Single-shot message
./target/debug/chatweb agent -m "Summarize today's tech news"

# Check configuration
./target/debug/chatweb status

# Install globally
cargo install --path .
chatweb agent

Deploy to AWS Lambda

# Prerequisites
brew install zig && cargo install cargo-zigbuild
rustup target add aarch64-unknown-linux-musl

# Build for Lambda ARM64 (must use musl, not gnu)
cargo zigbuild --manifest-path crates/nanobot-lambda/Cargo.toml \
  --release --target aarch64-unknown-linux-musl

# Or use the deploy script
LAMBDA_FUNCTION_NAME=nanobot-prod ./infra/deploy-fast.sh

Environment Variables

API Endpoints

System Requirements

Platforms: Linux (x86_64, ARM64), macOS (Apple Silicon, Intel), Windows (WSL2), AWS Lambda (ARM64)

Security

• Sandboxed execution -- tool code runs in isolated /tmp/sandbox/{session_id}/
• HMAC-SHA256 password hashing with configurable keys
• Rate limiting -- 5 login attempts/min, 3 registrations/min
• Webhook signature verification -- Telegram, Facebook, Stripe
• Audit logging -- 90-day TTL in DynamoDB
• CORS whitelist -- only configured origins allowed

See SECURITY.md for vulnerability reporting.

Roadmap

• Multi-model failover with circuit breakers
• Voice-first UI (STT + TTS)
• 6 channel integrations (Web, LINE, Telegram, Discord, Slack, Facebook)
• 50+ built-in tools with agentic loop
• Skill marketplace (publish and install custom tools)
• A/B testing framework
• Stripe billing integration
• Long-term memory engine
• SSE streaming
• WebSocket transport (Q2 2026)
• Multi-agent orchestration (Q2 2026)
• On-device LLM inference via GGUF (Q3 2026)

Project Structure

nanobot/
  crates/
    nanobot-core/         Core library: handlers, tools, providers, memory
    nanobot-lambda/       AWS Lambda entrypoint
  nanobot-cli/            CLI binary
  web/
    index.html            Web SPA (embedded into binary via include_str!)
    skill.html            Skill marketplace UI
    pricing.html          Pricing page
  infra/
    deploy-fast.sh        One-command Lambda deploy
    template.yaml         SAM template

Contributing

git clone https://github.com/YOUR_USERNAME/nanobot.git
cd nanobot
cargo test --all
cargo clippy --all-targets

See CONTRIBUTING.md for guidelines.

License

MIT -- Copyright (c) 2025-2026 nanobot contributors

Acknowledgments

• HKUDS/nanobot -- original Python nanobot (this project is a complete Rust rewrite)
• axum, tokio, serde -- the Rust ecosystem that makes this possible
• Anthropic, OpenAI, Google -- LLM providers

chatweb.ai -- voice-first AI assistant · teai.io -- developer API

Both powered by nanobot.