AI Chat

Bring your own LLM — local (Ollama, LM Studio, llama.cpp) or cloud (OpenRouter, Groq…) — and chat from your browser. 📖 User guide

🔒 Everything stays in your browser. Your endpoint, keys and chats are stored locally only, and requests go straight from your browser to your endpoint (or Google / Wikipedia / Jina for the search tools). Nothing touches our servers.

Settings, models & keys

Quick setup — choose a provider

Endpoint URL (OpenAI-compatible)

API key (optional for local servers)

Model

Compare models (check to send side-by-side)

Voice input

STT model

Read replies with

TTS model / voice

Embeddings model

Model prices JSON (per 1M tokens, for cost tracking) — e.g. {"gpt-4o":[2.5,10]}

Google Search (optional, 100 free queries/day)

Custom search engine URL (optional — SearXNG / 4get / any JSON API; use {q} for the query)

System prompt

Temperature

0.7

Max tokens

If your server runs on localhost, it must allow browser CORS: Ollama does by default; LM Studio and llama.cpp may need a --cors flag.

🧰 Tools & options

Enter sends (Shift+Enter for a new line) 🔊 Read replies aloud automatically Remember last messages in prompts

💡 Ask the model for fenced code blocks — each block gets a ⬇ download button (md, html, css, js, json, py, txt…). Replies also have ⬇ .md to save the whole answer.

🧠 Use saved memories in prompts 🛠️ Allow tool use (search, read, calculate, Python…) 📐 JSON replies (response_format) 🧲 Semantic memory (embeddings search) 👥 Compare with checked models 🧭 Expose custom search to the agent