4i.codes

4ICODE

FOR I, FOR ME

Made for you — the AI gateway built for developers

One domain, one key — access an expanding set of AI capabilities

Model services More capabilities 4ICODE Your code

Multiple models, one endpoint at 4ICODE

Get started See how to connect in one line

Official token pricing · FX advantage · transparent tier multiplier

Continuously expanding access · see the console for the live availability list

Text generationStructured outputTool callingLong context

Manifesto

Tokens uncut, Invoices unpadded.

Tokens uncut. Invoices unpadded.

The easiest thing for a reseller to do is quietly tweak things: truncate context, swap in a smaller model, hide something in the bill. You can't see it, but your usage remembers. We've closed that path — requests forward as-is, usage returns as-is, and the final amount is itemized as official token price × real FX discount × tier multiplier. This isn't a discount — it's a public formula.

Upstream usage returned prompt 1,842 / completion 906 Passed through as-is
Official token price input $1.25 / output $10.00 (per 1M) Aligned
Final cost Official × FX discount × tier multiplier Transparent & auditable

Every request log can be verified line-by-line by request_id in the console. No guessing, no asking support why it's cheaper than official — the formula is right here.

10k+ Developers served
Live Live availability in console
0 30-day uptime
0 Gateway median latency

Proof

Every promise, backed by data.

0%Markup on official pricing
100%Line-item reconciliation
∞Balance never expires
“Invoices finally reconcile line-by-line — finance stopped chasing me.”— Zhou Ning · Startup Tech Lead
1¥≈1$FX advantage
LiveLive availability list
99.97%30-day uptime
“Switched by changing one base_url — migration was nearly zero-cost.”— Chen Mo · Full-stack Developer
86msGateway median latency

Quick Start

Change one base_url —
your existing code stays untouched.

Fully compatible with OpenAI / Anthropic official protocols. Your SDKs, your frameworks, your agents — all work as before.

# Only change this line
export OPENAI_BASE_URL="https://api.4i.codes/v1"
export OPENAI_API_KEY="sk-4i-xxxxxxxx"

curl $OPENAI_BASE_URL/chat/completions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"MODEL_ID_FROM_CONSOLE","messages":[{"role":"user","content":"hi"}]}'
curl https://api.4i.codes/v1/messages \
  -H "x-api-key: sk-4i-xxxxxxxx" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{"model":"MODEL_ID_FROM_CONSOLE","max_tokens":1024,
       "messages":[{"role":"user","content":"hi"}]}'
Onboard · start in three steps

No architecture changes, no card required. From creating a key to your first response — at most three steps.

  1. 01

    Get a key

    Create an API key in the console, grouped by project with per-key limits and instant revocation.

  2. 02

    Point to 4i

    Point base_url to api.4i.codes/v1. Everything else stays the same.

  3. 03

    Start calling

    Every call appears in live logs instantly — latency, usage and cost, line by line.

01

Faithful

Route the model specified in each request as-is, with no silent downgrades or substitutions.

02

Uncut

Context forwards as-is — no truncation, no compression, no hidden prompt injection.

03

Auditable

Every request can be verified by request_id for usage and cost, with CSV export.

04

Resilient

Multi-node redundancy with auto-failover — requests reroute automatically when a node trips.

05

Restrained

No monthly fee, no forced plans, no expiry wipe. Your balance is yours — keep it as long as you like.

06

Private

We don't store request or response content — only billing-required metadata.

Use cases

What you can do with it.

One endpoint, from a single chat message to a production workflow.

Chat & assistants

Add streaming chat, multi-turn context and function calling — call 4i like you would the official API.

Content generation

Copy, summaries, translation, structured extraction — one call, many output formats.

Agents & tools

Long context + tool calls + multi-step reasoning — orchestrate models into automated pipelines.

Benchmarking

Run the same prompt across models, benchmark on one interface, compare at a glance.

Observability

Every request,
traceable.

Latency, token usage, upstream status, per-call cost — all live in the console. Billing is no longer a puzzle revealed at month's end.

Launch console
4i.codes · Console Live
99.97%Uptime 30d
86msP50 latency
1.24MRequests today
200gpt-5.6-sol12ms1,842 tok$0.0271
200gpt-5.6-terra8ms3,210 tok$0.0192
200gpt-5.6-luna15ms644 tok$0.0096
200gpt-5.3-spark6ms2,091 tok$0.0125
Your usage report, whenever you want it
Every request, every cost — export CSV reconciliation from the console in one click. See a sample invoice.
View sample invoice

Infrastructure

Multi-node relay —
critical calls never fall behind.

Requests auto-select the optimal path; when upstream jitters or throttles, the gateway reroutes to backup nodes in milliseconds instead of throwing errors at your users.

  • Automatic retry & reroute on upstream 429 / 5xx
  • Failed requests are never billed
  • Streaming responses pass through end-to-end, never fully buffered
us-west eu-central ap-east sg-01 jp-01

Transparent Billing

Price isn't our leverage —
it's our promise.

We thrive on scale and engineering efficiency, not on multipliers you can't understand.

Markup · never hidden in model prices

Commissions · no monthly fee, balance never expires

Auditable · verify line-by-line by request_id

Aligned with official Official token pricing

Input, output, cache reads and cache writes are all priced at upstream official rates — no markup on model prices.

1¥≈1$ FX advantage

Better exchange rates convert USD pricing to CNY spend, instead of billing at the official site's rate.

0.x Tier multiplier

Each tier sets a published multiplier (usually 0.x). Formula: final cost = official price × FX discount × tier multiplier.

100% Auditable

Line-item logs + CSV export — verify usage and price sources against official bills by request_id.

Same usage — how much do 12 months of costs differ?

Illustrative: 4i at official rates vs a reseller with markup (assuming steady monthly usage).

4i · official rates Typical reseller · with markup
4i example ¥100 Reseller example ¥130

Example only illustrates the no-hidden-multiplier gap — not a real bill; actual costs as shown in the console.

Myths, busted

About resellers, some things need to be said.

01

Markup isn't inevitable

Official token price × published multiplier — nothing hidden in model prices.

02

Commissions aren't required

No monthly fee, balance never expires — your money is always yours.

03

Black boxes aren't the norm

Line-item request_id reconciliation, published formulas — no guessing.

04

Cheap ≠ swapped

Faithful forwarding of the requested model, with no silent downgrades.

05

Reconciliation without support tickets

CSV export + live logs — verify every line yourself.

About the low price: your lower final cost comes from better FX rates and published multipliers — not acquisition bait like “free credit”. If there's ever a bonus, it shows up in your console.

FOR I, FOR ME

This gateway is built
the way we wanted to use it ourselves.

If you also believe faithful forwarding and honest billing should be the default, we're probably the same kind of person.

Get started Read the docs

Official token pricing · FX advantage · transparent tier multiplier

Voices

What developers say

Placeholder copy — replace with real authorized quotes before launch.

We load-tested faithful forwarding: requests route to what they specify, with no silent substitution.

— Lin Chuan · AI Infrastructure Engineer

Invoices match the official bills — finance finally stopped bothering me.

— Zhou Ning · Startup Tech Lead

Changed one base_url and we were in — the team migrated almost for free.

— Chen Mo · Full-stack Developer

FAQ

Before you start, you might ask

Availability changes and expands over time. Refer to the live list in the console and the API docs for the current scope. One domain and one key access everything currently enabled.
Final cost = official token price × FX discount × published tier multiplier. The formula is public — nothing hidden in model prices, no hidden markup.
Requests route to the model specified, with no silent substitution or downgrade. Each response can be checked by request_id for source and usage, and the sample quote covers forwarding consistency load tests.
We don't store your request or response content — only billing-required metadata, which you can export and delete from the console at any time.
No monthly fee, no minimum spend, balance never expires. That's a product promise, not acquisition bait — any bonus shows up in your console.
Create an API key, point your base_url at 4i, change one line of config. Your existing code migrates with almost no changes.