Prompsit API
Prompsit API is a REST translation service from the Spanish language-technology firm Prompsit. It routes text and document translation across Apertium, AltLang and OPUS-MT engines, and adds quality scoring, corpus annotation and translation memory under a flat time-based subscription.
What is Prompsit API?
Prompsit API is the productised machine translation service of Prompsit Language Engineering, S.L., a language-technology company based in Elche, in the province of Alicante, and a spin-off of the Transducens research group at the Universitat d'Alacant. It is a versioned REST API served from edge.prompsit.com, with a public OpenAPI 3.1.0 document, forty endpoints under a /v1/ prefix and bearer-token authentication.
The service describes itself as an orchestrator rather than a single model. Requests are routed to an engine the client selects explicitly: Apertium for rule-based translation of regional European pairs, AltLang for conversion between language varieties, and Mozilla and OPUS-MT running on CTranslate2 for neural translation. Engine metadata is returned with the response, which is the point the vendor makes against general-purpose competitors. Twenty languages are covered, including English, German, French, Spanish, Portuguese, Italian, Dutch, Polish, Russian, Arabic, Chinese, Japanese, Korean and Ukrainian, alongside the less commonly served Catalan, Galician, Danish, Northern Sami and both Norwegian written standards. Eleven regional variants are listed, from Brazilian and European Portuguese to three flavours of Valencian.
Beyond translation the same API carries the tooling a translation workflow usually has to assemble separately. Documents are handled as asynchronous jobs and cover Office formats, PDF to DOCX, XLIFF, PO, JSON, YAML, HTML, Markdown, CSV, TSV, subtitles, XML and plain text. Quality can be scored with BLEU, chrF++, TER, MetricX and COMET, either inline or from an uploaded file, and a separate reference-free endpoint checks whether tags and placeholders survived. Corpus scoring and annotation endpoints support data preparation, and a translation memory subsystem imports and exports TMX, edits segments and runs fuzzy search.
Access is developer-first and gated. A published npm package, prompsit-cli, provides a command line client; a device flow, a token endpoint and a Google OAuth callback handle authentication. Interactive documentation is open to anyone through Swagger and ReDoc, but credentials are obtained by submitting an email through a request-access form. Prompsit presents the product around privacy and predictability: short and configurable retention with secure deletion, request logs that avoid segment content, and a flat time-based subscription instead of per-token billing.
What it does
- Translate batches of text through one REST endpoint with explicit engine routing
- Translate Office, XLIFF, PO, subtitle and Markdown documents as asynchronous jobs
- Convert between language variants such as en-US and en-GB, pt-BR and pt-PT, or fr-CA and fr-FR
- Score translation quality with BLEU, chrF++, TER, MetricX and COMET
- Check that tags, placeholders and markup survive translation, without needing a reference
- Score and annotate parallel corpora as part of a data preparation pipeline
- Build and query translation memories, including TMX import, export and fuzzy search
When to use Prompsit API / When not to
A quick filter to help you decide if Prompsit API is the right fit.
When to use Prompsit API
- Localisation teams translating software strings, resource files and product copy at scale
- Back-end and full-stack developers embedding translation and document handling into their own product
- Language service providers that need Catalan, Valencian, Galician, Norwegian Nynorsk or Northern Sami coverage
- MT evaluation and MLOps engineers benchmarking engines with BLEU, chrF++, TER, MetricX and COMET
- Public-sector and regulated teams in healthcare, finance and legal that need short, configurable data retention
When not to use Prompsit API
- Individuals looking for a ready-made translation website or mobile app, since neither exists
- Buyers who need to see a published price list before contacting a vendor
- Teams needing broad Asian, African or Indic coverage beyond the twenty languages offered
- Non-technical users expecting a no-code connector for their CMS or translation management system
- Organisations that require a signed data processing agreement and a security certification before onboarding a processor
How to use Prompsit API
A typical end-to-end flow, from setup to results.
- Submit your email through the Request access form on the Prompsit API page
- Install the command line client with npm install -g prompsit-cli
- Authenticate with prompsit login, or request a bearer token from the /v1/auth/token endpoint
- Query /v1/translation/languages to see which pairs and engines are currently active
- Send a batch of text to /v1/translation, choosing the profile that routes to the engine you want
- For files, post to /v1/translation/document, then follow the job and download it from /result
- Import an existing TMX file into a translation memory and query it with fuzzy search
- Measure output quality with /v1/quality/score, or check tag preservation with /v1/quality/tags
- Prepare data by scoring and annotating corpora through the /v1/data endpoints
- Track consumption and plan limits with /v1/user/usage, and keep the Swagger documentation open while integrating
Pros & Cons
Pros
- Flat, time-based subscription announced instead of per-token billing, with limits framed as abuse control rather than metering
- Engine routing is explicit and documented, and the translation components are open source
- Real coverage of languages the large providers underserve, including Catalan, Valencian, Galician, Nynorsk and Northern Sami
- Quality evaluation, corpus scoring, annotation and translation memory sit inside the same API
- The full OpenAPI document and both Swagger and ReDoc are public, so a client can be generated before any contact
- The vendor is a twenty-year-old research spin-off involved in European projects such as OpenEuroLLM and HPLT
- Privacy positioning is concrete: short configurable retention, secure deletion and request logs that exclude segment content
Cons
- No price is published anywhere: no amount, no currency, no named tier and no pricing page
- Access is granted on request only, so there is no self-service sign-up and no way to evaluate cost beforehand
- No product terms of service exist, and the published legal documents cover only the corporate website
- Twenty languages is a narrow catalogue next to general-purpose translation platforms
- There is no web or mobile application, so every integration has to be written
- The retention and logging promises are marketing statements, backed by no contract, audit or published DPA
- No security certification is claimed or published, which is awkward for the regulated sectors the product targets
Pricing & Plans
No price is published. Prompsit states that the API is billed as a flat, time-based subscription rather than per-token or usage-based billing, and that rate and volume limits exist only to prevent abuse rather than to meter cost, each tier's limits being visible through the API itself. No amount, currency, tier name, free plan or free trial is announced anywhere on the site, and there is no pricing page: none appears in the sitemap, and the /pricing path simply returns the homepage. Credentials are obtained through a request-access form, so the entry ticket has to be established with the vendor directly.
Data, GDPR & hosting
A consolidated view of how Prompsit API handles your data.
GDPR overview
The privacy policy, last updated 16 September 2025, is formally GDPR-shaped. It names Prompsit Language Engineering, S.L. as controller, cites legal bases article by article (consent under 6(1)(a), legitimate interest under 6(1)(f), pre-contractual measures under 6(1)(b)), lists the rights of access, rectification, erasure, objection, restriction and portability, and points to the Spanish supervisory authority, the AEPD. It states that no automated decision-making or profiling with legal effects takes place, and that analytics data may reach Google LLC in the United States under the EU-US Data Privacy Framework, only with consent. The decisive limitation is scope: the document declares that it covers the website, not the API platform. The site claims to be EU AI Act and GDPR ready, but publishes no certification.
Who owns the data?
No product terms of service exist. The only published documents are a legal notice, a privacy policy and a cookie policy, all dated 16 September 2025, and the privacy policy states that its scope is the website prompsit.com rather than the API platform. Within that scope the controller is Prompsit Language Engineering, S.L. On the product page the vendor states that translation memories and corpora are stored only when the customer creates or imports them and remain under the customer's control, but that is a marketing claim, not a contractual one. Nothing published says who owns the text submitted for translation or the output returned.
Reuse rights
No contractual document says what the customer may reuse, and no permission regime can therefore be stated with confidence. In practice the API is designed for round-tripping the customer's own material: translation memories are imported and exported as TMX, corpus files are uploaded for scoring or annotation, and document results are downloaded per job. Reuse is restricted in the other direction. The legal notice forbids any commercial or promotional exploitation, reproduction, distribution or transformation of the website's own contents, including its texts, images, logos, code and domain name, without prior written authorisation. The machine translation components the service routes to are open source, but the API itself is declared Proprietary in its own OpenAPI document.
Data retention & training
Hosting summary
No hosting country or region is published for the platform. The privacy policy mentions hosting only for the corporate website, and only as a category: service providers that help operate the site, covering hosting and email, are referred to without being named. The single named international transfer is to Google LLC in the United States, for analytics, under the EU-US Data Privacy Framework and only if the visitor consents; the cookie policy adds that IP addresses from EU traffic are neither logged nor stored in the analytics tool. For the API itself the site says only that it is an API-first service on an edge endpoint, and that private deployment is possible for regulated teams, with on-premises or private cloud options offered on the services side of the business. No jurisdiction is committed to. What is certain is the governing law: the legal notice places the relationship under Spanish law, with the courts of Elche, Alicante, having exclusive jurisdiction, and the controller is established in Spain.
Things to keep in mind
Risks and trade-offs to weigh before adopting Prompsit API.
- No product terms of service exist, so the rules on liability, acceptable use and ownership of translated output are written nowhere
- The privacy policy declares that its scope is the website, not the API, which leaves the data you send to the platform contractually uncovered
- The short-retention and no-segment-logging promises are declarative: no data processing agreement, audit or certification supports them
- With no published price, a team can invest integration effort before discovering the commercial terms
- No ISO 27001, SOC 2 or equivalent certification is claimed, despite the product targeting regulated sectors
- Whether customer content is ever used to train models is never addressed, neither to confirm it nor to rule it out
- Machine translation of regulated content in healthcare, finance or law invites over-reliance: the quality metrics offered measure output, they do not replace human review
Setup & Integrations
Technical difficulty
Moderate, and clearly aimed at developers. You need an approved access request, Node.js if you want the command line client, and the ability to handle bearer tokens. There is no graphical interface and no ready-made connector, so any integration has to be coded, and document translation requires handling asynchronous jobs through polling or server-sent events. What lowers the effort is the documentation: the full OpenAPI 3.1.0 file, Swagger and ReDoc are all public, so a typed client can be generated before credentials are even granted.
Deployment
Supported languages
Behind Prompsit API
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Prompsit API.
Frequently asked questions
What can I build with the Prompsit API?
How do I get started?
Which languages does it support?
Which translation engines are used?
Which file formats can it translate?
How is customer text handled?
Which quality metrics are available?
How much does it cost?
Is there a mobile or web application?
Who is behind the product?
Should you pick Prompsit API?
Prompsit API is a credible and genuinely differentiated product rather than a repackaged wrapper around a large provider. The published OpenAPI document is authentic and detailed, the service reports a healthy status with a build dated the day of this review, and the endpoints describe a coherent workflow: translate, evaluate, score, annotate. Two things set it apart. The first is linguistic: Catalan and its Valencian varieties, Galician, both Norwegian standards and Northern Sami are properly served, which is rare. The second is architectural: engine routing is explicit and the quality and corpus tooling sits inside the same API instead of in a separate product.
The reservations are commercial and contractual rather than technical. No price is published at all, no tier is named, and access is granted only on request, so the cost of adoption cannot be assessed before a sales conversation. More seriously, there are no product terms of service: the three legal documents on the site are dated 16 September 2025 and the privacy policy states explicitly that its scope is the website, not the platform. Every reassuring statement about short retention, secure deletion and logs without segment content therefore rests on a product page, not on a contract, and no data processing agreement or security certification is published to back it. For a service that markets itself to healthcare, finance, legal and public-sector buyers, that gap is the main thing to raise before signing.
The publisher itself is solid: a 2006 university spin-off, active in European research programmes, with a long record of open-source language tooling. Prompsit API deserves a look from any team whose translation needs are European, regulated, or simply badly served by the large platforms.
- Choosing a selection results in a full page refresh.
- Opens in a new window.