Introduction
Automate your digital workflow through simple conversation.
Sparky is a desktop app that turns AI automation into a conversation. Connect agents, query your local documents, and integrate external services — all from a single, intuitive chat interface. Instead of stitching together pipelines and tools, you describe what you want and your agent handles the rest.
Whether you're automating repetitive workflows, synthesising research, or coordinating multi-step tasks across services, Sparky gives you the power of a full agent platform without the complexity of building one.
Why Sparky?
There is no shortage of AI chat apps and agents — but most live in the terminal. Sparky provides a simple yet powerful UI agent harness. Bring your own agent, and Sparky gives you the platform to manage your digital workflows with ease.
- Multi-provider agents — Use Claude, OpenAI ChatGPT, GitHub Copilot, Google Gemini, Google AI Studio, Mistral or run models locally with Ollama. Switch providers per chat.
- Chat organizer — Flag, label, archive, and search your conversations. Pin important messages as anchors so they never fall out of context.
- Rich content — Markdown, sortable tables, syntax-highlighted code, Mermaid diagrams, KaTeX math, and interactive ECharts visualizations, all rendered inline.
- Local RAG system — Add documents, code, and URLs to a production-grade knowledge base. Hybrid BM25 + vector search with reranking, powered entirely by llama.cpp no external API calls.
- Service integrations — Connect external APIs or MCPs like GitHub, Gmail, Telegram, or Todoist in one step. Your agent calls them directly from chat, with user approval before every action.
- AI-first interface — Every part of the app can be controlled through conversation. Create labels, change settings, connect services, manage workspaces, all without leaving the chat.