What is primerouter

Every response ships its own audit trail.

primerouter is a unified model gateway for AI API consumers. One account, one API key, access to hundreds of models — together with verifiable, honest commitments: no model swapping, no silent quantization, no token inflation, no covert prompt retention.

One-line definition

primerouter = one OpenAI-compatible endpoint + multi-provider routing + crypto-native billing + provable audit trail.

A drop-in replacement for the multi-platform, multi-subscription status quo (ChatGPT Plus + Claude Pro + Gemini Advanced + a domestic Chinese provider), unified into a single call entrypoint and a single bill.

Who uses it

Role	Scenario
AI developers	Existing OpenAI SDK projects — change one line of base_url to call dozens of providers (Claude, Gemini, DeepSeek, Qwen, ...)
AI tool users	Drop a single key into Cursor / Cline / LobeChat / ChatBox / n8n; every model is available
End users	No need to subscribe to several AI services individually — chat with GPT-4o / Claude / DeepSeek straight in the browser
Small teams	Centralized quota, usage, and billing — all settled by real consumption
Procurement	Top up flexibly with fiat, cards, or crypto; pay per real token; no monthly fees, no bundles

How we differ

primerouter's edge is not a feature checklist — it's four public commitments, programmatically verifiable:

1. Model identity preserved — no swapping

Request qwen-max, get qwen-max. Every response carries a verifiable upstream signature (upstream SHA + trace-id). When an upstream is down you get a 503 — never a 9B model wearing a GPT-5 nametag.

2. Full weights, full context — no quantization

Context windows, output limits, reasoning modes, FP precision — all passed through as upstream defaults. We won't silently route to a quantized version to cut cost. If we can't deliver at full capacity, we tell you.

3. Bills match upstream byte-for-byte — no token inflation

We don't multiply token counts. We don't round up. The usage field is byte-identical to what upstream returned. Invoices link to the per-request upstream ledger — pull them anytime.

4. No silent prompt retention — privacy by architecture

No-log-by-default — configurable, visible in source. The privacy policy is a contract, not a marketing page.

At a glance

your apps ─┐
           │
Cursor ────┤      ┌──────────────────┐       ┌─────────────┐
Cline   ───┼─────▶│   primerouter    │──────▶│ OpenAI      │
LobeChat ──┤      │                  │       │ Anthropic   │
n8n     ───┤      │ • OpenAI-compat  │       │ Google      │
browser ───┘      │ • Multi-format   │       │ DeepSeek    │
                  │ • Audit trail    │       │ Qwen, GLM…  │
                  │ • Crypto billing │       └─────────────┘
                  └──────────────────┘

Next

• Core features — everything in one page
• Architecture — how a request flows through
• 5-minute start — sign up to first call