Kelivo is a new open-source chat client built with Flutter that aims to be a single interface for every major LLM provider. The project, available on GitHub under the AGPL-3.0 license, supports both mobile and desktop platforms and includes features that go beyond basic chat, including multimodal input, web search integration, and tool calling through the Model Context Protocol.
The project is still in active development, with a TestFlight build available for beta testing on iOS. It supports Android, iOS, Harmony OS, Windows, macOS, and Linux. The breadth of platform support is notable for a chat client. Most LLM interfaces target either mobile or desktop, not both, and almost none support Harmony OS at all.
What Kelivo actually does
At its core, Kelivo connects to any LLM provider that exposes an API. OpenAI, Google Gemini, and Anthropic are explicitly listed, but the custom HTTP request headers and bodies feature means you can point it at any compatible endpoint. You configure providers once, and the app handles the rest.
The interface uses Material You design language with dynamic color theming on Android 12 and later. Dark mode is built in, and the app supports both English and Chinese interfaces. Custom fonts are available through system fonts, local imports, or on-demand downloads from Google Fonts. These are cosmetic details, but they matter for a tool you spend hours staring at.
Custom assistants let you create and manage personalized AI configurations. You can define system prompts, set model preferences, and save these as reusable profiles. This is useful if you use different assistants for different tasks, such as coding help versus writing assistance versus research.
Multimodal input and document handling
Kelivo accepts more than just text. You can send images, PDFs, Word documents, and other text-based files directly into the chat. The app renders responses with full Markdown support, including code syntax highlighting, LaTeX formulas, and tables. For developers, the ability to paste a code snippet or share a PDF and get a formatted response without switching tools is a real workflow improvement.
The voice features extend beyond basic text-to-speech. Kelivo integrates with OpenAI, Google Gemini, and ElevenLabs voice servers, plus it includes a built-in system TTS option. This gives you flexibility in how responses are read back, which matters for accessibility and for hands-free workflows.
Web search across 16 engines
One of the more unusual features is the breadth of web search integration. Kelivo connects to Bing, DuckDuckGo, Exa, Tavily, Zhipu, LinkUp, Brave, Metaso, SearXNG, Ollama, Jina, Perplexity, Bocha, Serper, and Grok. That is 16 search engines from a single interface.
For developers building with LLMs, this matters because different search engines return different results, and some are better suited to specific use cases. Exa and Tavily are optimized for AI workflows. Brave and DuckDuckGo prioritize privacy. SearXNG can be self-hosted. Having all of them available through one interface lets you compare results and pick the best source for a given query without switching tools.
The MCP integration is the other developer-facing feature worth highlighting. Model Context Protocol is a standard for connecting LLMs to external tools and data sources. Kelivo includes a built-in MCP Fetch tool, and the framework supports custom MCP tool integration. This means you can extend the chat client with your own tools, connecting it to APIs, databases, or internal services that the model can query during a conversation.
Practical features that matter
Prompt variables let you inject dynamic content into your messages. You can reference the model name, current time, or other contextual data without typing it manually each time. This is a small feature that saves time if you use structured prompts or need to timestamp your queries.
QR code sharing for provider configurations is a nice touch for teams. Instead of manually entering API keys and endpoint URLs on each device, you generate a QR code on one device and scan it on another. The configuration transfers instantly. For people who switch between phone and desktop, this eliminates the friction of reconfiguring providers on every device.
Data backup and restoration means your chat history is not locked to a single device. You can export your conversations and import them elsewhere. Combined with the QR code sharing for provider configs, this makes switching devices or setting up a new installation straightforward.
Android background generation is an optional setting that keeps the chat generation running even when you switch to another app. This addresses a real pain point on mobile, where the OS often kills background processes. If you send a long prompt and need to check something else, the response keeps generating instead of stopping when you leave the app.
Where Kelivo fits in the landscape
The LLM chat client space is crowded. Open-source projects like Chatbox, LobeChat, and others offer similar multi-provider support. Kelivo differentiates through platform breadth, the combination of MCP integration with web search, and the Flutter-based cross-platform approach that keeps the codebase unified across mobile and desktop.
The project credits RikkaHub for UI design inspiration, and the interface reflects that influence. It is clean and functional without being cluttered. The AGPL-3.0 license means you can self-host or modify the code, though the copyleft requirement means any modifications must be shared under the same license.
For developers who want a single client for interacting with multiple LLM providers, especially one that works on both phone and desktop and includes tool integration through MCP, Kelivo is worth watching. The TestFlight beta is available now for iOS testing, and the project welcomes pull requests and issues on GitHub.