LexiChat connects to your data — APIs, files, and MCP servers — and runs on the model you choose: 100% private with local Ollama, or any OpenAI-compatible provider when you want a frontier model.
A complete AI workspace that lives on your hardware.
Most assistants forget everything the moment you close the chat. LexiChat keeps a personal wiki — its long-term memory. Tell it something worth keeping and it files the fact away on its own; ask about it in a brand-new conversation and it searches the wiki, reads the page back, and picks up right where you left off. Perfect for todo lists, project notes, people, dates, and preferences that should outlive a single chat.
The wiki is just plain Markdown files on your own disk — no database, no cloud. Open them in any editor, edit them by hand, back them up, or delete them. They're yours.
Read the docs arrow_forwardLet Lexi write and run Python to compute exact answers — calculations, data processing, parsing and logic — instead of estimating. Every run happens in an isolated interpreter with no network and no surprise file access. You approve execution before it starts, and code can only touch the folders in your sandbox.
Ask Lexi to draw or photograph anything and it appears right in the chat — rendered entirely on your own machine with a built-in Stable Diffusion engine. No account, no cloud, no per-image cost. Pick a model (fast, or photorealistic) from a one-click picker; the engine ships inside the app. For photos, Lexi automatically writes a proper photographic prompt so results look real.
Generated locally · RealVisXL · 1024×1024
Run entirely on your own machine with Ollama and nothing ever leaves your device — that's the default, and it's completely private. Or connect any OpenAI-compatible endpoint when you want a frontier model: OpenAI, Anthropic, Groq, OpenRouter, Together, Mistral, or your own local server (LM Studio, llama.cpp, vLLM).
You stay in control of what's private. Add as many servers as you like and every model shows up in one list, tagged by where it runs. Pick a local model and the chat is 100% private; point a profile at a cloud provider and only that chat uses your key — nothing is ever sent anywhere you didn't choose.
Run local with Ollama and your conversations, files, and queries never leave your machine — no keys, no usage logging, no third-party servers. Prefer a cloud model? Bring your own key; only the chats you point at it are ever sent, and always to the endpoint you chose.
Read, write, search, and summarise files directly. Native support for PDF, Word (.docx), plain text, and code files. Browse your folders interactively in the chat.
Ask about current events, weather, or anything live. LexiChat searches DuckDuckGo and returns real results with clickable links — no browser switching needed.
Create multiple AI personas — a Code Assistant, a Writing Coach, a Research Agent — each with its own model, tools, system prompt, and file sandbox.
Connect any external service by pasting an OpenAPI spec, querying SPARQL / linked-data endpoints, or attaching MCP servers — which can even render interactive apps (live UI) right in the chat. Turn LexiChat into a powerful automation hub for your workflow.
Progressive chat parameters let you switch between Precise, Balanced, and Creative modes in one click — or dive into advanced controls like temperature, top-p, context window, and more. Set global defaults or change them per-chat.
Define prompts that run automatically — daily, weekly, every N hours, or on demand. Each job is tied to a specific profile and its tools, so it always runs with the right context regardless of which profile you have active.
Use the Step Builder to design multi-step workflows: chain text instructions with specific tool calls (web search, file read, datetime, and more) into a precise, repeatable sequence your agent follows every time.
LexiChat lives in your menu bar / system tray. Closing the window hides it — your scheduled jobs keep running. Click the icon to bring it back any time.
One-click profiles that turn your private local model into a domain analyst — wired to real open-data APIs, ready to research.
Cross-references the official UK Parliament APIs — members, questions, committees, interests — and cites every source.
Maps scholarly literature and the citation graph across OpenAlex and Crossref — find, trace, and cross-check papers.
Enter a postcode for a sourced neighbourhood profile — crime, house prices, planning, and local stats.
Chat interface with tool results
Settings & profile management
Extensible tools panel
Step Builder — design multi-step workflows with chained tool calls
Three steps. No account required. No API keys needed for local models.
For a fully private setup, download Ollama and pull a model — ollama pull gemma3 is a great, fast starting point. Prefer a frontier model? You can instead point LexiChat at OpenAI, Anthropic, OpenRouter, or any OpenAI-compatible endpoint with your own API key. Local keeps everything on your machine; a cloud key trades some privacy for a bigger model — your call.
Download the installer for your platform from the link above. macOS (Universal), Windows, and Linux (AppImage + .deb) all supported.
Open LexiChat, select your model, and start a conversation. Add folders to the Sandbox to give Lexi access to your files, or connect external APIs via the Tools panel.
LexiChat is free and always will be. If it saves you time or sparks joy, consider supporting continued development.
Send sats directly — no platform fees.
Every bit of support helps keep LexiChat free and actively maintained.
Questions or feedback? Reach us at info@lexi-chat.com