An AI chat workspace where you pick the right model for the task

Kotonia's AI Chat lets you switch between lightweight Gemini Flash and Qwen / Llama / Gemma / GPT-OSS / Nemotron / MiniMax in one screen. Whether you're coding, rewriting prose, translating, or studying, pick the model that fits. Tool integration and ReAct agent mode are available on the free tier.

10+ modelsOpen-weight switching

Free tier10–20 calls/day

Tools + AgentFunction calling / ReAct

Multi-model switching

Swap between Gemini Flash, Qwen, Llama, Gemma, GPT-OSS, Nemotron, and MiniMax with one click — pick the right tool for the task.

Tool integration (Function calling)

Create and call custom functions from AI. Add web search, calculators, API integrations, or any capability you build.

ReAct agent mode

Handle multi-step tasks autonomously — run code, generate files, and complete research with one prompt. Free tier: Gemini 3.1 Flash-Lite, 10/day. Top-tier models (Claude / GPT / Gemini Pro, etc.) on paid plans.

Persistent sessions

Save chat history per session and continue conversations with full context. Run multiple parallel topics in separate sessions.

Coding assistanceWriting & editingTranslation & summaryLearning supportIdea brainstormingResearch automation

Basic usage

Pick a model

Default is Gemini 3.1 Flash-Lite (the only model available on free tier). Switch to other open-weight models (Qwen, Llama, Gemma, GPT-OSS, DeepSeek V4, etc.) from the header — these require a paid plan.

Send a message

Works as a normal chat. Sessions auto-save your history so you can pick up where you left off.

Sample conversation

The real strength is the seamless integration: AI chat works in the same screen as custom tools, the ReAct agent, and the Linux runtime. Below is a single conversation flowing through model switching → tool search → ReAct execution → adding a new tool.

You

Web-search the latest best practices for parallel async execution in Rust.

Model: Gemini Flash + web-search tool

Crawled the official docs, Hacker News, and r/rust via a custom tool. Since 1.40, `tokio::task::JoinSet` is the recommended pattern. Want a concrete code example?

Function calling (custom tool)

You

Yes — switch to Qwen 3 Coder, write a sample, and actually run it to show me the result.

Switched model to Qwen 3 Coder. Spun up a Linux sandbox via ReAct: generated → executed → formatted result in 3 steps. 3 parallel tasks finished in 312ms ↓

Multi-model (paid) · ReAct · Linux runtime (coming soon)

You

Nice. I'd also like to add a tool that queries our internal DB.

Register the function name, args, and implementation (HTTP call works) in the custom tools screen. After that, both regular chat and the ReAct agent can call it automatically. Let me draft a template for you.

Custom tool definition / extensible

Available models (excerpt)

A subset of available models. **Only Gemini 3.1 Flash-Lite is available on the free tier** — the rest are delivered via OpenRouter and require a paid plan.

Gemini 3.1 Flash-LiteDeepSeek V4 FlashDeepSeek V4 ProQwen 3 CoderLlama 3.3 70BGemma 4 31BGPT-OSS 120BNemotron 3 Super 120BMiniMax M2.5

Pairs with other features

AI Chat connects seamlessly with the rest of Kotonia. Combine with tool integration and the ReAct agent to go from simple chat to complex task automation — all in one conversation.

Custom tools (Function calling)

Register functions the AI can call — web search, internal APIs, calculators. The AI calls them automatically when relevant.

Learn about tools →

ReAct agent mode

Hand off multi-step tasks (run code, investigate, generate files) with one prompt — the agent plans, executes, and finishes the job. Free tier available (Flash-Lite, 10/day).

Learn about the agent →

Multi-model comparison

PREMIUM

Run the same prompt on Claude / GPT / Gemini Pro / DeepSeek V4 in parallel and compare side by side.

Learn about multi-model →

Public personas

Pick a public persona built by other users with one click and use it as your chat partner.

Learn about personas →

Try free AI Chat now

One-minute signup, no credit card. Multi-model and tool calling included on the free tier.

Start free View pricing