Client de bureau léger et performant, multiplateforme : chat IA + AI Agent + passerelle IA
User requested a system architecture flow and token migration mockup. Structuring 4 stages of pipeline.
Clean layout, fluid workflow — multi-model routing, local knowledge base, and gateway status at a glance.
Available for macOS (Apple Silicon / Intel), Windows x64, and Linux distros. Fast native startup.
Plug in your OpenAI, Claude, DeepSeek API keys, or connect Ollama. Keys stay local and encrypted.
Multi-model parallel chat, MCP tools, sqlite-vec knowledge base, and local 127.0.0.1:8080 gateway.
Unified multi-model chat, autonomous Agent approval, local API gateway, and vector knowledge base in one native desktop app.
Connect 50+ providers including OpenAI, Claude, Gemini, DeepSeek, and local Ollama. Compare models side-by-side.
Runs at 127.0.0.1:8080 with OpenAI-compatible endpoints. Share keys & models directly with Claude Code, Codex, and Cursor.
Executes file edits and terminal commands with workspace sandboxing and explicit user confirmation dialogs.
Official Model Context Protocol support. Connect database connectors, terminal utilities, and external APIs directly.
In-process vector embeddings for PDF, Markdown, and codebase documents. Complete privacy with zero cloud indexing.
Configure system instructions, sampling temperatures, and import curated personas from community prompt markets.
Built-in support for LaTeX formulas, Mermaid diagrams, D2 architectures, Monaco code editor, and interactive Artifacts.
API keys stay on your machine encrypted with AES-256. Optional backup snapshots to WebDAV, S3, or local disks.
Free and open-source. Local-first privacy, multi-model chat, autonomous Agent sandbox, and built-in API gateway.
