Launching today
Polycode
One prompt. Every model. One answer - with receipts.
0 followers
One prompt. Every model. One answer - with receipts.
0 followers
Polycode asks Claude, GPT, Gemini, Apple Intelligence and your local models the same question at once, then hands you one synthesized answer — plus the full trace of who said what, how long it took, and what it cost. Citation pills flag where the models disagree, so you can judge confidence instead of trusting one black box. Native macOS. Your own API keys in Keychain, straight to providers — no proxy. Pay once, no subscription. Launch month: $19.95 → $4.95 through Aug 31.
Hi Product Hunt — I'm Anthony, and I built Polycode.
The problem I kept hitting: I'd ask Claude something, get an answer, and have no idea whether it was right. So I'd re-ask GPT. Then Gemini. Then sit there manually diffing three answers across three tabs, with no real sense of where they actually disagreed or what I'd just spent.
Polycode does that automatically. One prompt fans out to every model you've configured — Claude, GPT, Gemini, Apple Intelligence, or a local Ollama/LM Studio endpoint — they answer in parallel, and a primary model you pick synthesizes them into one reply. Citation pills mark the spots where the models diverged, so disagreement becomes a signal instead of something you never find out about.
The part I care most about is the trace inspector. Every exchange keeps its receipts: per-model latency, input/output tokens, resolved model ID, and cost. Nothing hidden, nothing estimated.
It's a native SwiftUI app, sandboxed, using your own API keys out of the macOS Keychain and talking straight to providers — no proxy, no account, no server of mine in the middle. Conversations stay on your Mac. One-time purchase, no subscription.
For launch month it's $4.95 instead of $19.95, through Monday August 31.
What I'd most like feedback on: does the consensus synthesis actually earn its place, or would you rather see the raw model responses side by side and judge for yourself? I went back and forth on that for months and I'm still not certain I picked right. Happy to get into the fan-out architecture or the cost accounting with anyone who's interested.