Your work,
amplified.

Tap once and just talk. Memo AI listens via Groq Whisper-large-v3-turbo (150ms transcription), replies out loud through Cloudflare MeloTTS, then listens again. Walk around, drive, multitask — the conversation continues until you stop.

Upload contracts, policies, PDFs. Each document is chunked, embedded with Gemini text-embedding-004 and stored in pgvector. Every chat silently retrieves the top-5 most relevant passages and answers from them. Per-user, 50 documents max.

When the AI generates a long code block or HTML doc, an "Open in canvas" pill appears. A slide-in side panel renders the code with copy, download and (for HTML) a sandboxed live preview. The same paradigm as Claude Artifacts and ChatGPT Canvas.

When the AI gives you Python, hit Run. Pyodide loads on demand (~5s first time, instant after), executes in a browser sandbox, prints output inline. Auto-installs numpy / pandas / scipy / matplotlib on import. No server. No Docker. No install.

Give it a complex goal — "plan a team offsite for 12 in Manchester", "draft a job spec + interview rubric + sample questions". A planner decomposes the goal into 3–5 steps, workers execute in parallel, a synthesiser combines outputs into one polished answer. Streamed step-by-step.

Memo AI learns about you — your name, your department, your manager, your writing preferences — by extracting facts after every conversation and injecting them into the next. View, add or delete memories at any time. Full memory settings page included.