Open-source toolkit for building and self-hosting voice AI agents with Rust services and ESP32-based hardware.
Best suited to
Builders experimenting with a self-hosted voice AI device
Open-source toolkit for building and self-hosting voice AI agents with Rust services and ESP32-based hardware.
Read the setup guide, configure the speech and model services, and test the voice device in your own environment.
Access and pricing: The server repository is GPL-3.0. Hardware, hosting, and model or speech providers can cost extra; open-source software does not include free inference.
The EchoKit server repository describes the server as the central component linking an EchoKit device to AI services. You configure it in a config.toml file, choosing services for voice activity detection, speech recognition, the language model and text-to-speech, plus prompts and MCP servers. It accepts OpenAI-compatible endpoints for speech recognition, language models and voices, with built-in pipelines for Google Gemini and Alibaba's Qwen Real-Time.
Clone the repo, edit config.toml, build with cargo build --release and start the server. Test with the chat page at echokit.dev/chat, then set up a device over Bluetooth at echokit.dev/setup by entering WiFi details and the server's WebSocket URL.
The server is GPL-3.0. Hardware is either the EchoKit device or a custom ESP32 running the firmware, and you also pay for any cloud model or voice services you connect. On aarch64 the build may need an extra flag. Comfort with Rust tooling and config files helps.
Voice agent server
Rust implementation
ESP32 device integration
Self-hosted setup
The server repository is GPL-3.0. Hardware, hosting, and model or speech providers can cost extra; open-source software does not include free inference.
Plan information can change. Prices and conditions are set by the provider. Information checked 2026-10-06.
Source for access detailsOne email a week with pricing and feature changes in this category. No per-tool price alerts.
By subscribing you agree to receive this email from Genius Forges and to us storing your address to send it. We may ask you to confirm by email. Every email has an unsubscribe link. See our privacy policy.
Read the setup guide, configure the speech and model services, and test the voice device in your own environment.
The server repository is GPL-3.0. Hardware, hosting, and model or speech providers can cost extra; open-source software does not include free inference.
Official-source information was reviewed on 2026-10-06. Plans and limits can change; check the linked provider source before subscribing.
The EchoKit server is open source under GPL-3.0, so the software is free. You still pay for hardware, hosting and any cloud speech or language services you connect.
EchoKit is used to build and self-host voice AI agents. The server connects an EchoKit device to speech recognition, an LLM and text-to-speech services that you configure.
Per the repository, you can use the EchoKit device sold through echokit.dev or a custom ESP32 device running the EchoKit firmware. The server itself runs locally or on your own host.
The README says any OpenAI-compatible API works for speech recognition, language models and voices, with built-in pipelines for Google Gemini and Qwen Real-Time, and ElevenLabs listed for streaming voice.
Tell us what needs improving. No account or private output required.
Report incorrect information or a broken linkBackend for coding agents: model access, user login, Postgres and file storage under one project key, with a $0 starter.
Best for: Vectoree is for developers and teams whose coding agents, such as Cursor, Claude Code or Codex, build apps that need a backend.
Open-source local guardrail that checks AI coding agent tool calls, secret reads, MCP use and package installs before they run.
Best for: HOL Guard is for developers who let agents such as Claude Code, Codex or Cursor run commands on their machine.
Free 14-minute unassisted coding diagnostic for JavaScript, TypeScript and Python, with paid practice drills for debugging and tracing.
Best for: Unrust is for developers who want to measure their own unassisted coding after leaning on AI assistants.