- Category
- AI Agents
- Rank
- No. 198Tools index
- Pricing
- Open Source
- Type
- AGENT
- Builder
- hkuds
- GitHub
- 47.9k stars
- Latest release
- v0.3.0
- Date
About
Lightweight open-source AI agent that works across your tools, chats, and workflows — integrates with Claude, OpenAI Codex, and OpenClaw.
What it does
Nanobot is a self-hosted Python runtime for personal agents. Messages enter through its browser workbench, terminal, API, or supported chat services, then share one agent loop that selects a model, runs approved tools, and returns the result. It preserves sessions, consolidates useful history into long-term memory, schedules work, delegates tasks, and connects external tool servers.
Why it's ranked here
Nanobot is compelling because its pieces form a coherent operating environment, not merely a chat wrapper. The same runtime serves local conversations, messaging channels, scheduled turns, persistent memory, and programmatic clients. Workspace controls, visible tool activity, model presets, and restorable memory changes address practical operation. The cost is meaningful setup and security ownership, especially when exposing channels or powerful local tools.
What's good
The browser interface covers model setup, topic management, workspace selection, access modes, tool activity, diffs, integrations, and automations. Memory has distinct short-term, archived, and durable layers, with version history and restoration. Provider presets support hosted services, local compatible servers, OAuth accounts, and fallback routing. Restricted workspaces, pairing, WebSocket tokens, sensitive-field blocking, and localhost-first browser setup provide concrete security boundaries.
Tradeoffs
It requires Python 3.11 or newer, while source installation also needs Bun or npm to build the browser interface. Users must supply and correctly match provider credentials, model names, and endpoints. Chat connections and scheduled work require a continuously running gateway. Remote exposure adds token, proxy, allowlist, and platform-permission work. Local triggers deliberately exclude webhook hosting, third-party authentication, and event formatting.
How to use it well
Use Nanobot when you want one self-hosted agent to retain context, work across projects, accept messages from several surfaces, and run scheduled or long-lived tasks. Start locally in the browser, verify one model, select a restricted workspace, then add capabilities individually. It suits operators comfortable owning credentials and gateway security. It does not replace an external webhook service or solve model procurement and pricing decisions.
Technical notes+
pyproject.toml defines the nanobot-ai Hatchling package, requires Python 3.11+, registers nanobot.cli.commands:app as the CLI entry point, bundles Python modules, skills, templates, channel assets, and built WebUI artifacts, and configures strict BasedPyright plus a 75 percent coverage floor. README.md documents PyPI, uv, pipx, source, macOS/Linux, and Windows installation paths. docs/concepts.md describes the shared agent-loop and gateway architecture. docs/memory.md specifies cursor-based JSONL consolidation, Dream updates, and Git-backed restoration. docs/websocket.md documents streaming JSON frames, multiplexed chats, TLS, allowlists, static or issued tokens, and trusted-proxy authentication.
Observed
- License
- MIT
- Primary language
- Python
- Runtime requirement
- Python 3.11 or newer
- Packaging
- Hatchling package published as nanobot-ai, installable with pip, uv, pipx, or from source
- Interfaces
- CLI, browser WebUI, Python SDK, OpenAI-compatible HTTP API, and WebSocket protocol
- Messaging channels
- Documents Telegram, Discord, Slack, WeChat, Email, Mattermost, Feishu, WhatsApp, QQ, Matrix, Microsoft Teams, Signal, and others
- Platform installation
- One-command installers are documented for macOS, Linux, and Windows PowerShell
Read from README.md, pyproject.toml, docs/webui.md, docs/README.md, docs/memory.md, docs/my-tool.md, docs/concepts.md, docs/chat-apps.md, docs/providers.md, docs/websocket.md.
What it can do
Generate code from natural language prompts
Natural language description of desired code → Generated source code
Integrate AI assistance across multiple tools
User's existing software tools and workflows → AI-enhanced tool functionality
Process chat conversations with AI models
User messages and conversation context → AI-generated responses
Execute automated workflows with AI guidance
Workflow definitions and trigger conditions → Completed automated tasks
Connect to Claude AI for advanced reasoning
User queries requiring complex analysis → Claude-powered insights and responses
Interface with OpenAI Codex for programming tasks
Programming requests and code context → Code suggestions and implementations
Utilize OpenClaw for specialized AI operations
Task requirements compatible with OpenClaw → OpenClaw-processed results
Tags
Tech Stack
Media
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.