Relay365 relay service for heavy Claude Code users
A heavy Claude Code day, told twice: the juggling first
Imagine one ordinary working day for a developer who lives inside Claude Code. The morning starts with three browser tabs open: one vendor dashboard for one model family, another for a second, a third for a side lane used only when the main one is busy. Keys are copied from one place and pasted into another. A rate that looked fine last month has quietly changed, and nobody remembers which tab holds the current number. By lunchtime the real work has not started, because the morning went to bookkeeping.
This is the version of the day most relay users know. It is not a complaint about any single provider; it is simply what happens when every lane sits on a different table. Each extra vendor adds a login, a minimum top-up, a separate invoice and a separate place to look when a request fails. The mental load is the hidden cost: before writing one line of code, the developer is already working a second job as an accounts clerk.
Relay365 was built as the opposite of that picture: a single duty desk where the lanes meet, run like a round-the-clock duty desk so the developer does not have to be.
The same afternoon from one table at the Relay365 duty desk
Now tell the same day again, this time through Relay365. The developer opens one dashboard, picks a lane, and pastes one base URL into the client. Every model on the desk, from Claude and GPT to Gemini, DeepSeek, Qwen, Grok and the rest, is reached from the same place, with the same key format and the same log. There is no second tab, because there is no second vendor to check.
Relay365 is an independent third-party relay, and it behaves like a duty desk: requests come in, get routed to the chosen lane, and get logged. Claude models run only from the official Claude Code and Claude Desktop clients, on Claude Code 2.1.285 or newer; other clients pair with the non-Claude lanes, and Gemini runs on its native protocol through Gemini CLI. The desk does not ask you to learn a new workflow per model; it asks once, at setup, with quick-wiring snippets and API keys that carry tags and bind to a lane.
The dashboard itself is calm: six interface languages, in-site messages to the desk, a ledger that separates wallet, bounty and pass accounts, and a pricing page that shows the full table. One table is not a slogan here; it is the layout of the screen.
Four token types, each priced on its own line
Heavy Claude Code users watch tokens the way dispatchers watch fuel. Relay365 prices the four token types separately on every lane: input, output, cache write and cache read each carry their own rate, so a cache-heavy workload is charged differently from a fresh-prompt workload, honestly and visibly.
Take Claude Opus 5 (wire name claude-opus-5) as an example. On Relay365 the rates per 1 million tokens are input $2.02, output $10.07, cache write $2.48 and cache read $0.1236; the official list price is input $5 and output $25, so the Relay365 output rate is about 40% of official. For example, if you send 1 million output tokens a day, Relay365 charges $10.07 against $25 at list, a difference of $14.93 for that single day, worked out only from these rates. Claude Sonnet 5.5 (wire name claude-sonnet-5-5) tells a similar story: output $5.91 against an official list output of $10, about 59% of official, with input $1.18, cache write $1.48 and cache read $0.0944 per 1 million tokens. The desk never hides a token type inside a blended number.
When a duty pass pays off
For steady, heavy usage there is an optional duty pass for the Claude and GPT lanes. A pass comes with its own dedicated key that never touches the wallet, which makes budgeting simple: the pass quota is drawn at the duty-pass rate, and the wallet only sees whatever falls outside it.
Two passes suit the heavy Claude Code user. The Claude 14-day pass costs $50.97 for 14 days and gives up to $700 of quota drawn at the duty-pass rate x2.5, with at most $380 of quota per 7 days. The Claude 30-day $3200 pass costs $216.25 for 30 days, gives up to $3200 of quota drawn at the duty-pass rate x2.5, at most about $1280 of usage at list prices, and allows at most $900 of quota per 7 days. If your day looks like the juggling version above, with long output-heavy sessions on a frontier Claude model, the pass is the line item that turns a scattered spend into one predictable number per month.
The weekly ceilings matter too: they keep a pass from being drained in the first weekend, and the dedicated key means pass traffic is never mixed into wallet billing. Other model families always use the wallet, so the desk stays transparent about which pocket paid for which call.
Every request traceable in the usage log
Trust at a duty desk is built on receipts. The Relay365 dashboard keeps a usage log of the latest 200 requests, showing the time, model, input, output, cache recall, charge and result of each call. If a stream disconnects, it is billed for what was actually used; if a call fails upstream with a 502 or 503, it is not billed at all. Billing is a prepaid wallet with no monthly fee, and every call is settled on the token counts the upstream reports.
The edges are published plainly as well. More than 120 calls in 60 seconds on one key, or more than 8 requests in flight for one account, returns HTTP 429. New accounts receive a $0.25 signup credit. Within 14 days of a top-up, an unused balance can be refunded minus what was consumed and the payment channel's fees; bounty and promotional credit are not refundable. Referrals pay a bounty of 18% at level 1 and 6% at level 2 of what referred members pay, withdrawable from $11.11 or movable into the wallet.
Put the desk to work on your next working day
If the juggling version sounds like your Tuesday, the single-table version is one signup away. Create a Relay365 account, grab a key, point Claude Code at the desk, and let the usage log tell the day's story from one table.