MCP Servers · 78

xiaozhi-esp32

xiaozhi-esp32 is an open-source AI chatbot firmware for ESP32 microcontrollers that integrates large language models (Qwen, DeepSeek) with voice interaction, speaker recognition, and IoT device control via MCP protocol. It supports 70+ hardware platforms and includes offline wake-word detection, multi-language support, and both device-side and cloud-side extensibility.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/78/xiaozhi-esp32

27.9k

GitHub stars

6.3k

Forks

C++

Primary language

MIT

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	78/xiaozhi-esp32
Owner	78
Primary language	C++
License	MIT — OSI-approved
Stars	27.9k
Forks	6.3k
Open issues	654
Latest release	v2.2.6 (2026-04-19)
Last updated	2026-07-07
Source	https://github.com/78/xiaozhi-esp32

What xiaozhi-esp32 is

C++ firmware leveraging ESP-IDF 5.4+, built on streaming ASR+LLM+TTS architecture with OPUS audio codec, offline ESP-SR voice wake-up, and dual communication protocols (WebSocket/MQTT+UDP). Targets ESP32-C3, ESP32-S3, and ESP32-P4; implements MCP for IoT control and integrates speaker recognition via 3D-Speaker model.

Quickstart

Get the xiaozhi-esp32 source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/78/xiaozhi-esp32.git2cd xiaozhi-esp323# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Voice-enabled IoT edge devices

Deploy AI chatbots on resource-constrained microcontrollers to enable natural language control of smart home devices, lights, servos, and GPIO without cloud dependency for offline wake-word and local decision-making.

Embedded AI prototyping and education

Quickly prototype AI-powered hardware on 70+ supported ESP32 boards using pre-built firmware and web-based asset customization, with comprehensive documentation for DIY developers and academic research.

Multi-modal intelligent peripherals

Build custom wearable or handheld AI assistants with speaker recognition, emoji display, battery management, and cloud-side MCP for extended capabilities (smart home, email, knowledge retrieval, desktop automation).

Implementation considerations

Set up Linux-based development environment with ESP-IDF 5.4+ and VSCode/Cursor to minimize compilation delays and driver issues on Windows.
Choose target hardware from 70+ supported boards or create custom board definition per [Custom Board Guide](docs/custom-board.md); test partition table compatibility before production builds.
Configure large model backend (Qwen/DeepSeek) via xiaozhi.me console or self-host using community server implementations (Python/Java/Go); plan for API key and LLM service costs.
Plan offline vs. cloud mode: offline wake-word uses ESP-SR and device-side MCP; full capabilities (smart home, email, desktop ops) require cloud-side MCP integration.
Validate audio pipeline (OPUS codec, ASR+TTS latency, speaker recognition accuracy) on target hardware; battery and power management configuration depends on form factor.

When to avoid it — and what to weigh

High-volume production without custom integration — If you need a drop-in commercial chatbot solution with guaranteed SLAs and vendor support, this open-source project lacks professional maintenance contracts and is community-driven.
Strict data privacy with cloud-mandatory setup — Default configuration connects to xiaozhi.me server; offline-only deployments require self-hosting community-maintained servers (Python/Java/Go), adding operational complexity.
Cross-platform mobile or web UI as primary target — Project focuses on ESP32 firmware and IoT control. While related client projects exist (Android, Python, Linux), the core is embedded; unsuitable if your main need is web/mobile frontend development.
Legacy hardware or older ESP-IDF versions — Requires ESP-IDF 5.4 or above; v1→v2 upgrade incompatibility means partition table rework. Older ESP32 variants or development boards not listed may lack tested support.

License & commercial use

MIT License (permissive, OSI-compliant). Allows free use, modification, and redistribution for any purpose, including commercial applications, with minimal restrictions (retain license and copyright notice).

MIT license permits commercial use without restriction, but: (1) this is community-maintained open-source with no vendor support, SLAs, or liability guarantees; (2) server backend deployment requires self-hosting or reliance on community-maintained implementations; (3) dependencies (ESP-IDF, Qwen/DeepSeek APIs, third-party audio codecs) may have separate commercial licensing terms—review each carefully before shipping.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Adequate
License clarity	Clear
Deployment complexity	Moderate
DEV.co fit	Good
Assessment confidence	High

Security considerations

No security audit or formal vulnerability disclosure process mentioned. Considerations: (1) firmware runs on-device with local voice processing and WiFi/4G connectivity—ensure network isolation per your threat model; (2) default cloud connection to xiaozhi.me; audit that service's TLS, authentication, and data handling; (3) speaker recognition and audio data handling—verify codec and storage do not leak sensitive voice data; (4) OTA firmware updates—ensure signed/verified deployment pipeline if used in production; (5) MCP device control can execute arbitrary commands—validate and authorize all endpoints; (6) open-source nature means community review but no professional security maintenance.

Alternatives to consider

Rhasspy (open-source voice assistant)

Full offline voice assistant stack with ASR/TTS/NLU; runs on Raspberry Pi and x86. Better for pure software integration; lacks embedded microcontroller optimization and hardware diversity.

Mycroft AI (voice assistant platform)

Modular voice assistant with skill ecosystem and cloud/on-device options. More polished UI and ecosystem but heavier resource footprint; less suited to resource-constrained ESP32.

Google Assistant SDK / Amazon Alexa Device Kit

Commercial off-the-shelf with professional support, large device ecosystem, and robust cloud backend. Requires licensing and cloud dependency; less hackable and higher cost.

Software development agency

Build on xiaozhi-esp32 with DEV.co software developers

Start with xiaozhi-esp32's pre-built firmware on xiaozhi.me, explore custom hardware integration via the developer docs, or contact us to plan a production deployment with professional support, backend hosting, or custom MCP extensibility.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

xiaozhi-esp32-server-java py-xiaozhi esp-ai

Related on DEV.co

Explore the category and the services that help you build with it.

MCP Servers AI Application Development IoT & Custom Hardware Integration Cloud Deployment (for self-hosted backend)Talk to DEV.co

xiaozhi-esp32 FAQ

Can I use xiaozhi-esp32 without connecting to xiaozhi.me?

Partially. Offline features (wake-word detection via ESP-SR, device-side GPIO/LED control via MCP) work without cloud. However, full LLM chat, cloud-side device control, and advanced features require either xiaozhi.me or self-hosting one of the community server implementations (Python/Java/Go).

Is v1 still supported? Can I upgrade v1 firmware to v2?

v1 (stable 1.9.2) is maintained until February 2026. Direct OTA upgrade from v1 to v2 is not supported due to partition table incompatibility; manual firmware reflash is required. Switch to v1 branch via `git checkout v1` if upgrading is not feasible.

What hardware do I need to get started?

Any of 70+ supported ESP32 boards (e.g., Espressif ESP32-S3-BOX3, M5Stack CoreS3, LiChuang ESP32-S3). Beginners can use pre-built firmware binaries flashed via web interface at xiaozhi.me. Development requires Linux + ESP-IDF 5.4+ for custom builds.

How do I integrate custom IoT devices (smart home, PC control)?

Via MCP protocol: device-side MCP handles local GPIO/speaker/LED; cloud-side MCP extends capabilities. Consult [MCP Protocol IoT Control Usage](docs/mcp-usage.md) and [MCP Protocol Interaction Flow](docs/mcp-protocol.md). Requires backend implementation and LLM prompt engineering.

Software development & web development with DEV.co

Need help beyond evaluating xiaozhi-esp32? DEV.co is a software development agency offering software development services and web development for teams of every size. Our software developers and web developers build custom software, web applications, APIs, and mcp servers integrations — and maintain them long-term.

Ready to Build AI-Powered IoT Devices?

Talk to DEV.co Schedule a Consultation