One gateway. Every model.
A unified AI gateway with OpenAI- and Claude-compatible APIs, distributed nodes that auto-pick the lowest latency, and pure pay-as-you-go pricing — top up and spend only what you use.
How it works
Issue
Create a scoped API key for each app, teammate, or agent. Revoke one key without touching the rest.
Route
Point any OpenAI- or Claude-compatible SDK at Tokenswim. Distributed nodes automatically pick the lowest-latency path for each request.
- OpenAI protocol
- Claude Messages
- Chat Completions
- Responses API
Meter
Pay only for tokens you use. Usage rolls up per key, model, and day — with logs when something looks off.
Every request should take the fastest path.
Usage that explains itself
Rollups per key, model, and day. See who used what before the bill tells you.
Scoped keys
One key per teammate or agent. Revoke it in a click, the rest keep swimming.
Request logs
Latency, model and token counts for every call, searchable when something looks off.
Quotas and controls
Per-key budgets and access controls. Spend what you need — nothing is bundled into seats or unused plans.
Help power the network.
Capacity partners contribute routing capacity to Tokenswim's distributed gateway. Applications are reviewed, credentials stay encrypted server-side, and traffic is metered end to end.
Many models in. One key out.
- claude code
- codex
- gemini cli
- grok
- web app
- ci bot
- research
One endpoint. Zero rewrites.
The gateway speaks OpenAI (Responses and Chat Completions) and Claude (Anthropic Messages). Swap the base URL in any SDK, keep one key, and every request lands in your logs.
curl https://tokenswim.app/v1/responses \
-H "Authorization: Bearer $TOKENSWIM_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gpt-5.6-sol",
"input": "Hello"
}'Questions, answered.
What is Tokenswim?
Tokenswim is a unified AI gateway: one API key, one base URL, and access to modern LLM models through OpenAI-compatible and Claude-compatible endpoints. Distributed edge nodes route each request toward the lowest-latency available path.
Is Tokenswim OpenAI- and Claude-compatible?
Yes. Swap the base URL in any OpenAI SDK for Chat Completions or the Responses API, or use the Anthropic Messages shape at /v1/messages. Keep your existing client code — only the endpoint and key change.
How does pricing work?
Pure pay-as-you-go. Top up credits when you need them and spend only what you use — no monthly plan, seat minimum, or bundled commitment. Live per-model rates are on the Pricing page and at /pricing.json.
How does routing choose a node?
Traffic is served by a distributed node network. For each request the gateway prefers the available path with the lowest observed latency, so apps stay responsive as demand shifts across regions.
Can I contribute capacity to the network?
Yes — approved capacity partners can apply to contribute routing capacity under Tokenswim's partner program. Access is reviewed; credentials stay encrypted server-side and never appear in client apps.
How is Tokenswim different from other AI API relays?
Tokenswim is an AI API relay — point any OpenAI- or Claude-compatible client at it by changing the base URL. The difference is billing you can audit: input, output, cache-read and cache-write tokens are priced separately and published at /pricing.json, billing is pure pay-as-you-go with no monthly fee or seat minimum, every API key is revocable on its own, and per-key usage rollups plus request logs (latency, model, token counts) are inspectable per call.
Is Tokenswim a shared membership or an account-sharing service?
No. Shared memberships and co-renting mean several people take turns on one ChatGPT Plus or Claude Pro account and have to share login credentials. Tokenswim gives you your own API key and a token-metered balance: no shared password, no queueing for a seat, keys revocable individually, and usage and billing that belong only to you. It is for wiring models into your own app or CLI, not for signing in to a web chat UI.