Sente: The Open-Source Coding Assistant You Can Talk To

A Coding Assistant That Lets You Speak Instead of Type

Typing commands in a terminal or chatting in a text box — that's the standard way developers interact with AI coding assistants. Sente, a tool that recently appeared on Product Hunt, takes a different approach: it lets you simply talk to it. Built by Japanese developer Yuki Hamada, the tool centers on two-way voice interaction. You can use it in the terminal or pull it up on your phone at any time, making the whole experience feel as natural as talking to a colleague.

The product earned 63 upvotes on Product Hunt and ranked 18th, under the categories of Software Engineering, Developer Tools, and Artificial Intelligence. While it's not a viral hit, its approach to interaction and openness makes it worth a closer look.

Two-Way Voice: Register Your Voice in 10 Seconds

Sente's core selling point is "Voice, both ways" — true two-way voice interaction. It can understand what you say and read results back to you out loud. To enable this, you simply read a sentence aloud (about 10 seconds) to complete your voice enrollment.

The developer places particular emphasis on a "consent-first" principle: voice enrollment requires explicit user authorization, and the delete key is always yours. The product also makes a clear commitment — we never clone a voice that isn't yours. At a time when AI voice synthesis is increasingly being misused, putting privacy and consent front and center is a responsible signal.

For developers, the real value of voice interaction is freeing your hands and attention: issuing verbal commands to an agent while debugging, reading documentation, or moving around can be more fluid than switching windows to type. That said, how well voice holds up for precise code editing scenarios still needs more real-world feedback to verify.

Voice enrollment is technically known as "Voice Enrollment" or "Speaker Registration." The system extracts acoustic features (such as pitch, speech rate, formants, etc.) from a short phrase the user reads aloud, generating a voice vector used for subsequent speaker verification or personalized speech synthesis. A ~10-second enrollment window is considered a lightweight approach in the industry — sufficient to capture basic acoustic features, but less personalized than high-fidelity cloning solutions that require several minutes of recording. Sente's voice enrollment is most likely used for "Speaker Verification" or to drive customized TTS (text-to-speech) output, rather than full voice cloning. Understanding this distinction helps set realistic expectations for the product — it lets the system "recognize your voice," rather than generating a synthetic replica that sounds exactly like you.

One Account, Access to 404 Models

Another highlight is Sente's model aggregation capability. The product claims a single account gives you access to 404 models, covering major LLMs including Claude, GPT, Gemini, DeepSeek^Ad, and more. Crucially, it's compatible with both the OpenAI and Anthropic API specifications.

This compatibility means existing toolchains and scripts can be migrated over at low cost, without maintaining separate integrations for each model provider. For developers who need to compare models side by side or want to switch flexibly based on the task at hand, a unified entry point can significantly reduce management overhead.

Supporting both OpenAI and Anthropic API specs means Sente handles protocol adaptation at the interface layer. OpenAI's Chat Completions API has become a de facto industry standard, and many third-party model services (such as DeepSeek, Together AI, etc.) also use compatible interfaces; Anthropic's Messages API, meanwhile, has its own request structure and parameter system. By supporting both, Sente allows developers to migrate existing call code or prompt engineering scripts with minimal rewriting. For teams that have already configured model-calling logic in frameworks like LangChain or LlamaIndex, this compatibility can dramatically shorten onboarding time.

MIT Open Source, Built in Japan

Sente is open-sourced under the MIT license, with code hosted on GitHub at github.com/yukihamada/sente. MIT is one of the most permissive open-source licenses available, allowing free use, modification, and commercial use — which is quite developer-friendly for teams that want to self-host, build on top of it, or audit the codebase.

The product also wears its "Made in Japan" identity proudly: it's deployed in the Tokyo region and supports billing in Japanese Yen (JPY). For developers in Japan and the Asia-Pacific region, localized data residency and currency settlement are genuine conveniences, and also address data compliance concerns to some degree.

Data Residency is an increasingly important compliance factor for enterprises and developers when selecting cloud services — it refers to data being physically stored and processed within a specific geographic region. Japan's Act on the Protection of Personal Information (APPI) has explicit requirements for cross-border transfers of personal data, and certain industries (such as finance and healthcare) face even stricter regulatory constraints. Sente's choice to deploy on AWS Tokyo (ap-northeast-1) means requests and data are processed within Japan, helping meet local compliance requirements and reducing legal uncertainty around data transfers abroad. For developers or enterprise users in the Asia-Pacific region, this deployment decision carries practical value equal to the features themselves.

Honest Billing Limits

Unlike many products that advertise "unlimited" usage, Sente takes a candid approach to pricing. The official documentation states plainly: metered, not unlimited. Users receive a certain amount of credits each month and can top up when they run out.

This "honest limits" framing is itself a product philosophy — no vague language to create a gap between expectation and reality. Behind the access to 404 models lies real compute and API costs, and transparent metering is ultimately healthier for both sides, giving developers a clearer picture of what they're spending.

Takeaway

Sente brings three things together: natural voice interaction, a unified multi-model entry point, and open-source engineering with local deployment. It's not trying to out-feature mainstream coding assistants on breadth — it's carving out differentiated space through interaction style and openness. Whether voice can truly become a high-frequency interface for coding agents remains to be seen; but for developers who like to tinker and care about privacy consent and open-source control, this is an early-stage product worth keeping an eye on.