HeyHank
Not a chatbot — a full-time AI coworker, soon to be released as open source.
HeyHank is an AI coworker that works like a real employee: he reads and replies to emails, schedules appointments, makes phone calls over SIP, posts on social media, writes code, builds presentations, and generates images and video. When work piles up he hires his own AI agents that operate like a team — you give the brief, Hank delegates.
Communication
IMAP/SMTP for email, SIP for phone calls with realtime voice (Gemini Live), calendar integration and social media APIs. All in one interface.
Multi-session coding
Hank works on multiple projects in parallel — separate workspace per session, files and git state stay isolated.
Media generation
Images via Imagen / gpt-image / Nano Banana, video via Veo, audio via ElevenLabs. All through a unified call.
Hires his own team
Hank can hire specialized sub-agents (research, bookkeeping, travel booking) and delegate to them — you're CEO, Hank is the head of department.
Why HeyHank?
Today's AI tools are fragmented: ChatGPT for text, another tool for email, a third for calendar, then a phone bot. In the end you do the work yourself anyway because the copy-paste between tools costs more time than it saves. HeyHank flips that — every channel in one interface, one agent controlling all of it.
For that to work, HeyHank isn't a wrapper around a single LLM. More than 40 AI providers are integrated — you pick the best model per task, or let Hank decide.
Stack
Bun and TypeScript on the backend, custom state management for parallel sessions, FreeSWITCH for the SIP bridge, Gemini Live for the voice pipeline. Everything is designed to run locally on your own machine — or on your own VPS, if you want Hank reachable at night.
Open source and roadmap
HeyHank will be released under MIT at the public launch. The private beta is running now — if you want early access, sign up via heyhank.app or reach out by email. The detailed case study with screenshots and a demo video is in the references section.