An applied AI company shipping products and running research. Curtis AI Chat — the Obsidian plugin below — is the first.
Curtis AI is an applied AI company. We ship products that earn their place in your workflow, and run a research arm that pushes the field forward.
The Obsidian plugin below is our first shipped product — a polyglot AI chat with thirty-plus providers, agent tools, voice, and a research-grade memory system. The lab is in setup. More products are on the way.
AI calls 9 tools to read, create, and edit your vault notes.
AGENT·01Stream one prompt to 2 models in parallel, side-by-side.
ARENA·02Cursor-style rewrite with accept/reject modal + assignable hotkey.
DIFF·03Type @ in chat, fuzzy-search your vault, attach as context.
Whisper speech-to-text in, browser TTS out.
VOICE·05Fuzzy-matched picker across all conversations + messages.
SEARCH·06Download any conversation as .md via /export.
Edit or delete individual memory facts. No more append-only.
MEMORY·08Curtis AI Chat is the first product. Curtis AI Research is the lab arm — separate from the plugin, set up to pursue grant-funded work at the intersection of personal knowledge systems and applied AI. We're standing it up now. If you fund, research, or build in this space, we'd like to talk.
Drop main.js, manifest.json, styles.css from the latest release into .obsidian/plugins/curtis-ai-chat/. Enable under Settings → Community plugins.
Open Settings → Curtis AI Chat → Providers. Enable one, paste a key. Keys live in your OS keychain (Obsidian 1.13+).
Click the robot in the ribbon (or Ctrl+Shift+G), pick a model, type, hit Enter. That's it.
Offline-first? Install Ollama, run ollama pull qwen2.5:7b-instruct, enable the Ollama provider. Nothing leaves your machine.