One API, every model across providers.
Use one key for GPT, Claude, Gemini, DeepSeek, Grok, and the rest of the live model catalog. Connect through the documented API and pay for measured usage.
Checking the latest Connect release…Built with teams moving AI forward.
From agent infrastructure to global payments, we work with teams that make the AI stack more useful in the real world.
LibreFang
Run autonomous agents through EveryAPI with native credential discovery and live model pricing.
Vaultka
A global virtual card platform supporting cryptocurrency, advertising, and subscriptions.
Routing— leave it to EveryAPI.
A production gateway needs routing policy, bounded retries, budget controls, metering, and usage records. EveryAPI provides those layers behind one API.
Same model, many routes — we select an eligible route automatically.
The relay filters channels by the requested model, account policy, health, priority, and configured weights. It selects among eligible channels; this is routing policy, not a guarantee that every model has redundant capacity or that the chosen route is always fastest.
For retryable failures, the relay can retry another eligible channel when one is available. Non-retryable errors, exhausted candidates, and responses that have already begun streaming may still fail.
Set limits before spend grows.
Set a monthly cap per key — once it's over budget, requests from that key are blocked automatically.
Every request is traceable.
Review metered usage by key, model, or time window. Each recorded request includes the usage and charge used for billing.
Choose the route that fits your network.
Use the global or China-priority documented endpoint; point your baseURL to the one that performs better in your own tests.
Account for cached tokens when the selected model reports it.
When an upstream model reports prompt-cache read or write tokens, EveryAPI records those token classes and applies that model's configured cache rates. Availability and savings depend on the selected model and provider.
One CLI line, or two env vars.
Claude Code, Codex, Cursor, and compatible SDKs can use EveryAPI. Either run everyapi use to connect in one line, or just change the API endpoint to api.everyapi.ai and add your key — both work.
For China-first routing, use https://api-cn.everyapi.ai/v1
- Use the official CLI directlyUse `everyapi use` for supported tools, or configure a documented API endpoint manually
- Desktop clients work tooDrop the baseURL into Cherry Studio / Chatbox / LobeChat and you're done
- One key for everythingA single key reaches the live model catalog through documented compatible APIs
1# 1. Install the everyapi CLI (macOS / Linux)2curl -fsSL "https://dl.everyapi.ai/install.sh" | bash34# 2. Sign in to bind your account (device-auth, once per machine)5everyapi auth login67# 3. One line to wire up official tools — env injected automatically8everyapi use claude # same for codex / gemini
1# Point the official Claude Code at EveryAPI's domestic nodes2export ANTHROPIC_BASE_URL="https://api.everyapi.ai"3export ANTHROPIC_AUTH_TOKEN="sk-everyapi-..."45# Then use it as usual — all Claude models go through EveryAPI6claude "Add type annotations to this code"
Where the money goes, all traceable.
View usage per key and per model, with balance, calls, and spend updating in real time. The model, token count, and charge for every request can be looked up.
Keys
Balance healthyThis month| Key | Model | Spend | Status |
|---|---|---|---|
| prod-claude-code | Claude Opus 5 | $24.80 | Active |
| team-codex | GPT-5.6 | $9.84 | Active |
| gemini-cli | Gemini 3.1 Pro Low | $8.15 | Active |
| ds-batch | MiniMax M3 | $5.96 | Active |
| test-key | Any | $0.60 | Paused |
One gateway, two sides of supply.
EveryAPI combines platform-operated supply with approved user channels behind the same caller-facing API. Routing, billing, revenue sharing, and risk controls are handled by the gateway; actual eligibility and availability depend on current configuration and channel health.
Deeper capacity, fuller niche coverage
Eligible platform-operated and user channels can serve the same gateway. Availability still depends on the requested model, channel health, account policy, and current capacity.
EveryAPI connects both sides.
We handle matching, risk control, billing, and anonymization so neither side is aware of the other. Both ends only ever talk to the same baseURL — whether the source switches or whose channel steps in, no one has to worry about it.
Share capacity without finding customers
After approval, an enabled and healthy channel can enter the eligible routing pool. Traffic and earnings depend on demand, model eligibility, routing policy, health, and configured limits; neither is guaranteed.
One key, a multi-provider catalog. The channel marketplace is the gateway's business.
- OpenAI-compatible API · baseURL = api.everyapi.ai, no code changes
- Visible model pricing · Usage is billed against the model prices shown in the live catalog
- Deeper capacity · Eligible platform and user channels share a pool; retryable failures may use another eligible channel
- Source-agnostic · The gateway schedules by weighted random — you don't have to choose
Plug in your idle keys, earn the configured revenue share, under the current dashboard rules.
- Bring a supported channel · Connect a supported credential type and complete platform verification
- Dashboard-visible revenue share · Into a separate wallet, transfer to main balance in one tap
- Automatic gateway scheduling · No pricing, no client negotiations, no support
- Less operational work · Health, rate limits, and settlement are handled by the platform
Pay what you top up, spend what you use.
Prepaid balance with usage-based billing. Per-model rates are loaded from the public pricing catalog; the authenticated dashboard is authoritative for your account's effective prices and permissions.
- Run a few demos to see results
- Compare 3-5 models
- Show a PoC to your boss
- Get $2 on signup, added to your top-up balance
- Balance never expires, save it for next time
- No card, runs until spent
- Daily Claude Code / Cursor use
- Ongoing calls for personal projects
- Running agents for research
- Cache hits billed automatically at hit price
- Isolate projects with multiple keys
- Usage breakdown by key / model
- Current model rates available in the dashboard
- Production calls
- Shared across a team
- Batch data / RAG retrieval
- Per-key spending limits
- Usage sliced by key / model / time window
- Monthly billing archives, fully auditable
- Retryable failures can use another eligible channel
- High-volume production workloads
- Multiple projects or environments
- Sustained agent and batch usage
- Per-key spending limits
- Usage breakdown by key, model, and time
- Itemized billing history in the dashboard
- Contact sales for account-specific arrangements
- 25 models available live
- Get $2 on signup, added to your top-up balance
- Balance never expires
- Requests stop when available balance is insufficient
- No card · no subscription
- Token-level real-time billing
| Model | Input | Output | Context |
|---|---|---|---|
| gpt-6-luna | $0.10 | $0.50 | — |
| claude-haiku-5-5 | $0.10 | $0.50 | 1M |
| glm-5.3-flash | $0.15 | $0.50 | — |
| gpt-5.6-luna | $0.20 | $1.20 | — |
| deepseek-v4-flash | $0.30 | $1.20 | 1M |
| deepseek-v4-pro | $0.30 | $1.20 | — |
Live public catalog · model-specific input, output, and reported cache-token rates · account dashboard is authoritative
Volume discounts · dedicated channel · account manager
Most people pick just one side, but both stay open.
Most people buying API don't have allowance to plug in, and most people with allowance to plug in don't need to buy API. The two sides are two kinds of users, but they enter the same account system. If you happen to do both, funds can move between them — that's all.
Your main balance drains one-way, exactly as before. No need to look at the other side at all.
Eligible channel earnings accumulate in the seller balance. You can transfer available earnings to the main account balance from the wallet when the account is not frozen.
Provide channels at peak, call the API in the lulls. What you earn from channels can offset next month's bill via Transfer.
Leave the rest to EveryAPI.
Get a key in 30 seconds, with $2 of free trial credit on signup. Try a few models first, then top up when it feels right.