AI Deck

Nitro API — A Translation API That Lets AI Agents Order Professional Human Translation Without an API Key

Nitro API is a translation API that lets AI agents order professional human translation autonomously, with no account registration and no API key issuance. It is designed for developers who want an agent to place external orders without holding any credentials, and payment is completed by responding to an HTTP 402 payment challenge. The translations themselves are done not by machines but by native-speaking professional translators, and the service is operated by the localization company Alconost Inc. This API is the ordering endpoint of “Nitro”, the on-demand translation service the company has been running since 2009, opened up to AI agents in July 2026.

Key Features

  • Order without registering in advance: An unauthenticated POST /v1/translate returns a 402 with a WWW-Authenticate: Payment challenge; resend the request with payment credentials attached and you get a 200 OK with an order ID. Because there is no API key to issue, store, or rotate, you can hand external ordering to an agent without giving it long-lived credentials
  • Pay-per-order via MPP: Payment follows MPP, an open standard jointly drafted by Stripe and Tempo. It is built around HTTP 402, exchanging WWW-Authenticate: Payment / Authorization: Payment headers, and the payment itself is settled between the agent and the MPP provider. The conventional prepaid-balance-plus-API-key model coexists with it, so you can choose from three schemes: Basic, Bearer, and Payment
  • Humans do the translating: Native-speaking professional translators handle each job, with QA and revision included as standard. Orders move through QUEUE → PROGRESS → DONE, and post-completion edits are tracked separately as REVISION. According to the official figures, turnaround averages 2 to 24 hours, with 63% of orders finished within 2 hours
  • Submit content with markup intact: Requests accept four resource types ─ plain text, HTML, JSON, and iOS strings (text/x-objcstrings) ─ so only the body text is translated, leaving tags and keys untouched. XML and CSV are supported only through the web file-upload flow. Note that while the official site advertises “80+” supported languages, the language-code list in the API documentation has 61 entries, so the two are counted differently
  • Specify tone and subject area: The optional context parameter lets you set the tone (FORMAL / INFORMAL / GUESS, which leaves it to the translator), a maximum character count for the translation, and a subject category. There are 13 categories, including legal, medical, finance, gaming, crypto, e-commerce, and IT
  • Get quotes without authentication: GET /v1/rates, which returns unit prices, and POST /v1/calculate, which estimates the cost before ordering, both require no authentication, so an agent can evaluate its budget before proceeding to place an order

Pricing

Billing is per 1,000 source characters, with no minimum order. Unit prices vary by language pair.

Language pairPer 1,000 source characters
English → Japanese$31.40
English → German$23.50
English → French$20.50
English → Chinese (Simplified)$19.90
English → Spanish (Spain)$16.60

There are two ways to use the service. The MPP flow requires no setup and settles each order individually; the conventional flow tops up an account balance and deducts from it via an API key.

Prices are actual values retrieved from GET /v1/rates as of August 2026 (all 211 pairs are published through that endpoint). Check the official pricing page for the latest rates.

Pros and Cons

Pros

  • No API key to issue, store, or rotate, so you can let an agent place external orders without holding credentials
  • Cultural nuance, domain terminology, and brand voice ─ hard to guarantee with machine translation ─ are covered by human translation with QA included
  • You can lock in the cost before ordering with rates and calculate, which makes it easy to build autonomous execution with a budget ceiling
  • The backend is a translation service that has been operating since 2009, with an existing translator network and track record (publicly stated as over 10 million words per year and 91.8% customer satisfaction)

⚠️ Cons

  • The official documentation explicitly states that MPP payments currently support only US-issued cards, so this flow is unavailable with, for example, Japanese cards (the API-key approach still works)
  • Because translation is done by humans, responses are not instant. It assumes asynchronous processing of 2 to 24 hours and is unsuited to real-time UI translation
  • Unit prices are more than an order of magnitude higher than machine translation APIs. Pushing large volumes of content through wholesale is hard to justify

Comparison with Similar Services

CriteriaNitro APIGengoDeepL API
Who translatesHuman (native translators)Human (translator platform)Machine translation
Registration / API keyNot required (MPP flow) / API key also availableRequiredRequired
Billing unit1,000 source charactersWordCharacter count
Turnaround2 to 24 hoursHours to daysInstant

Who It’s For

  • Developers who want to delegate even external tasks like translation to an autonomously running AI agent
  • Teams translating text in fields where mistranslation causes real harm, such as legal, medical, finance, and gaming
  • Anyone who needs to localize marketing copy, taglines, and other text that does not work when translated literally
  • Individuals and small teams who want to cover unpredictable translation demand with pay-as-you-go instead of a monthly contract

Summary

Nitro API takes a combination that was hard to reconcile ─ “human translation” and “an API that agents can order from autonomously” ─ and connects the two straightforwardly through HTTP 402 and MPP. An ordering flow with no key management pays off in situations where you want to avoid the risk of handing credentials to an agent. On the other hand, you have to accept that MPP-flow payments are limited to US-issued cards, and that human translation means asynchronous processing at a high unit price. A good starting point is to use the unauthenticated GET /v1/rates and POST /v1/calculate to check the unit price for the language pairs you care about.

← Blog