An OpenAI-compatible endpoint that routes one key across a live model catalog. On top of the router: virtual models that put two models on hard problems, and an agentic workspace with web search and GitHub context. Your first million tokens are free.
- import Chart from "react-chartjs"
+ const Chart = dynamic(() => import("react-chartjs"), { ssr: false })One OpenAI-compatible endpoint, a live model catalog behind it — and the things a plain router doesn't have: virtual models, web search, GitHub context.
OpenAI-compatible end to end. Change one line — your base URL — and every model in the live catalog is one API key away. Live per-model pricing at /api/v1/models.
Call dogecat/moa like any other model id — two models answer in parallel, then a synthesis pass merges the strongest parts into a single answer that beats either alone.
Toggle web search and the agent reads live results before it answers — current docs, changelogs, and issues instead of stale training data.
Paste any GitHub link — repo, file, or directory — and the agent reads it before replying. No setup, no indexing, no waiting.
Your task — plus any web or GitHub context — goes to two frontier models from the catalog at the same time.
An orchestrator pass reads both proposals: correcting errors, merging strengths, filling gaps.
One final answer, streamed live — metered cost-weighted, like any other call from your token pool.
One token pool, cost-weighted — each model draws at its own rate. Live per-model pricing at /api/v1/models.
1M free tokens on signup. The full workspace — dogecat/moa, web search, GitHub context.
Start freeFirst-party models from $0.20/M, in and out — cheaper ones draw less. $5/mo for 100M standard tokens — $100/mo for 2.5B — plus one-time top-ups.
View API docsMint credit-capped API keys for your own users. Per-key budgets, usage, and revocation — credit is standard tokens at the $0.20/M rate; models draw cost-weighted.
Platform keys docsNo credit card. The full workspace — dogecat/moa, web search, GitHub context — and every model in the catalog one OpenAI-compatible call away.