Multi-LLM collaboration
Every channel is a shared, multi-model session
Invite your team into a channel and @mention Claude, GPT, Codex, or Gemini by name. The whole thread is visible to everyone, so context never lives in one person's private chat history.
Shared by default
Channels are collaborative spaces — teammates read, reply, and prompt in the same thread.
Mix models freely
Use the right model for each turn. Mention more than one in a single conversation.
Real-time presence
See who's in the channel and follow the conversation as it unfolds.
/compare
One prompt, every model, side by side
Type /compare and HuddleSesh fans your prompt out to Claude, GPT, Codex, and Gemini at once. Read the answers next to each other and keep the one that fits.
Claude
GPT
Codex
Gemini
BYOK security
Bring your own key — we never resell AI
Connect one key per provider at the org level. HuddleSesh routes your prompts through your account, so you keep your own pricing and your own relationship with each provider.
Encrypted at rest
Keys are protected with KMS envelope encryption — never stored in plaintext.
Org-scoped
One key per provider per org. Reconnecting rotates the key; it never duplicates.
Server-side only
Keys are decrypted only on our server at invoke time — never shipped to the browser.
Never resold
We do not mark up or resell AI tokens. You pay your provider directly.
Observability feed
See progress without reading the code
A read-only, redacted live view for managers. Track an engineer's local AI-coding session at a glance — status, intent, and activity — while their private code stays private.
Status at a glance
Each session shows active, idle, blocked, or done — so you know who needs help.
Signal, not surveillance
Intent, branch, model, tokens, active minutes, and top files — redacted by design.
Read-only
Managers observe; they never see source contents or take control of a session.
macOS desktop app
A notarized native app for your Mac
Run HuddleSesh as a notarized macOS app on Apple Silicon. It powers the local AI-coding sessions that feed the observability feed.