D Route ships only what production AI needs. No filling pages with buzzwords to look impressive.
HTTP-level and SDK-level compatibility with the OpenAI Chat Completions spec. Swap the base URL, keep the client.
One credential, every model. No juggling seven access tokens across seven vendor dashboards.
Every frontier laboratory in one place — and 290+ specialised models you can switch per request.
Pass model: "auto" and D Route picks the best model for each prompt based on price, latency and capability.
Server-Sent Events streaming for chat and completions — first token in <400ms p50.
Send images, PDFs and screenshots. Multimodal understanding backed by GPT-4o, Claude 3.5, Gemini 2.0, Qwen-VL and more.
First-class support for OpenAI o1/o3, DeepSeek R1, Claude extended thinking — captured reasoning tokens included.
Function calling, parallel tool calls, structured JSON outputs and tool-use loops supported everywhere.
Tokens, latency, errors and cost — chunked per model, per key, per project, with daily charts.
Per-project API keys. Set spend caps, model allowlists, IP allowlists, expiration, and one-click revoke.
Edge-cached routing, smart pooling, and warm connections keep latency low for every region.
Create as many projects, teams and keys as you want. No per-seat fees, no caps.
Invite teammates, share keys safely, and audit every request — perfect for agencies and startups.
Tag requests with project IDs to see per-project spend, latency and errors instantly.
Each key can have its own model allowlist, rate limits, spend cap, allowed origins and webhook URL.
Test any prompt against any model in the browser before you write a line of code.
SOC2-aligned operations, encrypted at rest, TLS everywhere, audit logs for every request.
Email tickets answered in <24 hours. Plus an active Discord and Telegram for power users.