Direct upstream access, MA token billing, zero-migration OpenAI protocol — a decentralized AI model gateway built for developers.
curl https://macdecloud.com/v1/chat/completions \
-H "Authorization: Bearer dcld-sk-..." \
-d '{"model":"deepseek/deepseek-chat","messages":[{"role":"user","content":"Hello"}]}'
The first of TypeSafe's System One models. It does not write prose — given a decision to make, it returns a typed result your code can consume directly, with no parsing or fallback layer around it. Because the output is not long-form text, the output side is priced at zero: you pay only for what goes in ($0.042 / 1M), with a 32K context. It fits wherever your system has to judge something and then branch — intent detection, content triage, ticket routing, tool selection.
One API key connects to multiple model services through direct DeepSeek, Qwen, GLM and MiniMax channels, plus OpenRouter aggregation.
Multi-region coverage, low latency and high availability, first-token under 200ms.
Swap base_url and API key — existing OpenAI SDK code needs zero changes.
Generate a dcld-sk- key in one click — view usage, rotate, or revoke anytime.
On-chain billing via MA Chain — every request settled in milliseconds, balance fully transparent.
Direct DeepSeek, Qwen, GLM, MiniMax, with more frontier models added continuously.
Log in to the console with a supported wallet, transfer assets to the deposit address, and MA lands automatically on chain confirmation.
Generate a dcld-sk- key in one click — view usage, revoke, or rotate anytime.
Point base_url at the deCloud gateway and fire requests from any OpenAI-compatible SDK.