Skip to content

API documentation

With the Do4U API you use the same tools as on the website from your own software: summarise, compare, support replies, alt texts, rewrite and translate. Every request is processed on our own servers in Poland, EU.

Base URL: https://do4u.eu/api/v1

OpenAPI 3.1 description (JSON) – Machine-readable description of this API for code generators and tools such as Postman or Insomnia. No key needed.

Authentication

Create a key under API keys in your account. Send it with every request in the header Authorization: Bearer <key> (or X-API-Key: <key>). Keep the key secret and never put it into a web page. You can revoke a key at any time.

How it works

Jobs run asynchronously. You create a job and get its id at once (HTTP 202). Then you fetch the job until its status is done or failed. Ask every 2 to 5 seconds; most jobs take a few seconds.

Endpoints

GET /tools Tools, options and your input limits
GET /balance Your credit balance
POST /jobs Create a job (JSON, or multipart/form-data with an image file)
GET /jobs/{id} Status and result of a job
DELETE /key Revoke the key you send with this request, for example when an integration is disconnected. Jobs already created still finish.
POST /key/revoke The same as DELETE /key, for clients that can only send GET and POST.

Example

Create a job, then fetch the result:

curl -X POST https://do4u.eu/api/v1/jobs \
  -H "Authorization: Bearer d4u_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"tool": "summarize", "text": "...", "options": {"length": "short"}, "language": "en"}'
HTTP/1.1 202 Accepted
{"id": "8f1c0b6a...", "tool": "summarize", "status": "queued", "model": "do4u-standard",
 "credits_estimate": 1, "links": {"self": "https://do4u.eu/api/v1/jobs/8f1c0b6a..."}}
curl https://do4u.eu/api/v1/jobs/8f1c0b6a... -H "Authorization: Bearer d4u_YOUR_KEY"

{"id": "8f1c0b6a...", "tool": "summarize", "status": "done", "model": "do4u-standard", "credits_estimate": 1,
 "result": {"summary": "...", "key_points": ["...", "..."]}, "credits": 1,
 "notice": "AI-generated, please check before use."}
curl -X POST https://do4u.eu/api/v1/jobs -H "Authorization: Bearer d4u_YOUR_KEY" \
  -F tool=alt_text -F language=de -F "text=Product page: hiking boots" -F image=@photo.jpg

Request fields for POST /jobs

toolName of the tool, see the table below. Required.
textInput text. For compare: text A. For alt_text: optional page context.
text_bText B (compare only).
urlPublic web page to read instead of text (summarize, accessibility_check).
htmlHTML of a page to check (accessibility_check), up to 300,000 characters.
focus_keywordOptional focus keyword for seo_meta. Without target, seo_meta writes in language.
faqOptional FAQ snippets (support_reply only).
questionsOptional for seo_meta, content_brief and faq_schema: frequent visitor questions as a list of texts or {"text": ..., "count": n} (up to 50, at most 20 are used). faq_schema answers them first if the source answers them; visitor_questions in the result says what happened to each.
imageImage for alt_text: file field in multipart/form-data, or base64 / data URI in JSON. JPG, PNG, WebP or GIF, up to 5 MB.
image_urlOptional for alt_text instead of image: public address of the image (http or https; JPEG, PNG, WebP or GIF; at most 5 MB). Do4U loads it once in the background. If that fails, the job fails with error image_unavailable and costs nothing.
optionsObject with options of the tool, for example {"length": "short"}.
languageLanguage of the result and notes: en, de or pl. Default en. For translate, the target language is the option target.
no_storetrue: the input is deleted right after processing and the result after the first retrieval.
brand_voiceOptional: false switches your brand voice off for this job (default: the brand voice applies if you have set one).
ai_labelOptional: true shows the visible note "Marked as AI-generated" with the result in the web interface, false hides it. Default: the setting in your account. The machine-readable AI label is always in the result and cannot be switched off: the fields ai_generated and generator, an HTML comment in HTML results, ai_notice for image descriptions. ai_generated is false if the result comes from automatic checks only.
callback_urlOptional: public https address that gets a signed POST when the job is finished (see Webhooks).
render_jsOptional: true loads the page with JavaScript first (pages that build their content in the browser; tools that read a url). Costs the render surcharge only if the page was really loaded this way; the result field render says what happened.
no_cacheOptional: true always starts a new AI job. Without it, a finished and still stored result of your account (or workspace) with the same tool, options and input is reused: the job is done at once, from_history is true and it costs 0 % of the credits of the original. Never with no_store, never for a url.
allow_same_languageOptional for translate, translate_html and translate_file: true translates even if the text already is in the target language. Without it such a job is rejected with same_language (400) – only when the check is sure.

Repeat requests safely (Idempotency-Key)

Send the header Idempotency-Key with POST /jobs, for example a random UUID for each job (1 to 255 printable ASCII characters). If the connection breaks and you send the same request again with the same Idempotency-Key and the same API key within 24 hours, you get the same answer again: the same job, no second job and no second charge. The answer then has the header Idempotent-Replayed: true. Use a new key for every new job. If you send the same key with a different request, you get 422 idempotency_conflict. Errors are not stored: after an error you can send the request again with the same key.

curl -X POST https://do4u.eu/api/v1/jobs \
  -H "Authorization: Bearer d4u_YOUR_KEY" \
  -H "Idempotency-Key: 5f0c2a8e-7d1b-4c3a-9e2f-1b6d8a4c0e77" \
  -H "Content-Type: application/json" \
  -d '{"tool": "summarize", "text": "..."}'

Tools and options

ToolInputs Options (first value is the default)Result fieldsPrice
moderate_commentReview comments for spam text – classification, reason, suggestion, action, note usually 1–3 credits
document_htmlTurn documents into HTML image, pdf, text page: 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 | 50 html, plain_text, notes, page, note usually 2–3 credits
image_textFind baked-in image text url image_index: 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 html, plain_text, notes, images, image_index, coverage, note usually 2–3 credits
pdf_accessibilityCheck PDF accessibility pdf – checks, pages, truncated, repaired, note, convert_url usually 1–2 credits
summarizeSummarise text, url length: medium | short | long summary, key_points usually 1–3 credits
compareCompare text, text_b focus: overall | facts | tone | seo summary, similarities, differences usually 1–3 credits
support_replySupport reply text, faq tone: friendly | neutral | formal reply, concerns, questions usually 1–3 credits
alt_textAlt texts image, text, filename, image_url – alt, caption, description, filename, filename_base, extension, language, alt_length, alt_max, alt_status, caption_length, caption_max, caption_status usually 2–3 credits
rewriteRewrite text style: simpler | shorter | more_formal | more_casual | more_persuasive | easy_to_read text, changes, language, style_label, readability, easy_check, easy_note, review_status, review_limit, diff usually 1–4 credits
accessibility_checkAccessibility check url, html – summary, page, counts, findings, checks, readability, note fixed 3 credits per task
seo_metaSEO title & description url, text, focus_keyword, questions target: en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv variants, current, focus_keyword, limits, keywords, note fixed 1 credit per task
translate_htmlTranslate HTML html target: en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv html, segments, check, glossary, notes, note, review_status, alignment usually 1–4 credits
product_descriptionProduct descriptions text tone: neutral | friendly | persuasive | formallength: medium | short | longchannel: shop | marketplacetarget: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | svmeta_description: no | yes products, language, detected_language, channel, format, limits, counts, check, note usually 1–3 credits
headlinesHeadlines and subject lines text kind: page | newsletter | adtone: neutral | friendly | persuasive | formalcount: 5 | 3 | 8target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv kind, fields, variants, limits, counts, enough, report, language, detected_language, note fixed 1 credit per task
faq_schemaFAQ and schema markup url, text, questions type: faq | organization | localbusiness | productcount: 8 | 5 | 10target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv type, kind, jsonld, html, valid, faq, fields, checks, counts, unverified, dropped, existing, existing_errors, existing_same, page, language, detected_language, notes, note, visitor_questions fixed 2 credits per task
content_briefContent briefing text, audience, questions target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv title, title_length, title_max, title_status, goal, intent, outline, questions, length, markdown, counts, dropped, limits, language, detected_language, notes, note usually 1–2 credits
review_replyReplies to reviews text, company stars: 0 | 1 | 2 | 3 | 4 | 5platform: general | google | trustpilot | booking | tripadvisor | airbnb | yelp | app_store | google_playtone: friendly | neutral | formalcount: 3 | 2target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv assessment, variants, limits, platform, stars, tone, counts, dropped, enough, report, hints, checks, language, detected_language, note usually 1–2 credits
social_postsSocial posts from an article url, text channels: all | linkedin | facebook | instagram | linkedin+facebook | linkedin+instagram | facebook+instagramtone: friendly | neutral | persuasive | formal | casualvariants: 1 | 2 | 3target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv channels, posts, limits, counts, enough, report, link, link_in_bio, language, detected_language, source, notes, note usually 1–3 credits
explain_termsExplain terms url, text level: simple | general | expertexamples: no | yestarget: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv terms, level, language, detected_language, source, text_length, text_clipped, counts, dropped, adjusted, limits, other_candidates, checks, markdown, html, notes, note usually 1–3 credits
extract_dataExtract data url, text types: all | hours | prices | hours+prices | addresses | hours+addresses | prices+addresses | hours+prices+addresses | events | hours+events | prices+events | hours+prices+events | addresses+events | hours+addresses+events | prices+addresses+events | hours+prices+addresses+events | contact | hours+contact | prices+contact | hours+prices+contact | addresses+contact | hours+addresses+contact | prices+addresses+contact | hours+prices+addresses+contact | events+contact | hours+events+contact | prices+events+contact | hours+prices+events+contact | addresses+events+contact | hours+addresses+events+contact | prices+addresses+events+contacttarget: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv kinds, hours, prices, addresses, events, contact, counts, empty, unverified, dropped, limits, source, language, detected_language, notes, note usually 1–3 credits
redirect_planRedirect plan old_urls, new_urls, old_domain, new_domain ai: yes | no old_base, new_base, mappings, counts, unused_new, invalid, dropped, ai, legend, language, domain_ignored, notes, note usually 1–3 credits
contrast_checkColour contrast check colors level: aa | aaatext_size: normal | large colors, pairs, counts, invalid, settings, palette_check, legend, language, note usually 1–2 credits
language_levelLanguage level (A1–C1) text level: B1 | A1 | A2 | B2 | C1target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | svkeep_terms: no | yes text, changes, level, language, detected_language, translated, summary, within_guideline, measures, over_guideline, over_count, readability, content, kept_terms, checks, hints, note usually 1–4 credits
llms_txtllms.txt generator url target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv name, summary, details, file, filename, sections, counts, pages_read, read, existing, dropped, hints, language, detected_language, notes, note usually 1–3 credits
internal_linksInternal links text, pages, sitemap_url, page_url – links, text, text_is_html, counts, pages, dropped, ai, limits, language, detected_language, timeout, notes, note usually 1–3 credits
outdated_contentOutdated content url, text, html, reference_date target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv reference_date, reference_date_given, findings, summary, counts, source, dropped, ai, limits, language, detected_language, timeout, notes, note usually 1–3 credits
inclusive_languageInclusive and plain language text scope: both | exclusion | complexitygender: not_prescribed | neutral | pair | star | colon | brand_voice language, scope, gender, summary, counts, findings, candidates, measures, applied, changed, revised_text, diff, dropped, checks, hints, note usually 1–3 credits
review_analysisReview analysis text focus: all | criticismtarget: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv format, columns, language, detected_language, focus, counts, selection, stars, period, languages, languages_unknown, personal_data, themes, praise, criticism, page, faq, cited, dropped, limits, hints, notes, note usually 1–3 credits
ai_visibility_checkAI visibility check url, text, html, topic, questions target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv coverage, coverage_level, counts, questions, faq_questions, sections, qa_on_page, structured_data, llms_txt, robots_txt, readability, hints, source, topic, dropped, ai, limits, files, language, detected_language, note, how usually 1–3 credits
events_calendarEvents to calendar url, text, time_zone, year target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv events, counts, unverified, dropped, timezone, year, today, limits, ics, jsonld, filename, source, language, detected_language, notes, note usually 1–3 credits
product_photoProduct photo to text image, image_url, text, languages target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | svcategory_list: google | free language, detected_language, title, title_length, title_max, title_status, short_description, short_length, short_max, short_status, description, description_length, description_max, description_status, bullets, bullets_count, bullets_max_length, bullets_status, alt_text, alt_length, alt_max, alt_status, category_group, attributes, attribute_sources, attribute_list, translations, removed, hints, claims_checked, claims_unchecked, counts, limits, notes, category_list, jsonld, jsonld_check, product_json, check, note usually 2–3 credits
chart_descriptionDescribe charts and infographics image, image_url, text target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | svlength: normal | short chart_type, title, language, detected_language, alt, alt_length, alt_max, alt_status, description, description_length, description_max, description_status, description_size, axes, data, data_rows, table, data_csv, html, image_text, pie_check, hints, counts, limits, notes, verify_numbers, check, note usually 2–3 credits
design_feedbackDesign feedback from a screenshot image, image_url, text target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | svdevice: auto | phone | desktop language, detected_language, device, summary, findings, image, image_basis, counts, removed, hints, limits, notes, estimate, check, note, text, markdown usually 2–3 credits
job_ad_checkJob ad check url, text, html jurisdiction: auto | eu | de | pltarget: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv language, detected_language, jurisdiction, title, checklist, counts, pay, structure, length, findings, titles, summary, source, text_cut, dropped, ai, hints, note usually 1–3 credits
green_claims_checkGreen claims check url, text, html target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv findings, summary, counts, source, dropped, ai, limits, legal, language, detected_language, timeout, notes, note usually 1–3 credits
third_party_scanThird-party check url, html – url, host, page_from, static_only, counts, consent, groups, cookies, legend, notes, check, note, truncated, checked_at usually 1–2 credits
broken_link_checkBroken link check url resources: yes | noexternal: no | yes language, url, host, counts, broken, more_broken, redirects, more_redirects, soft, start_soft, suggestions, not_reached, limits, note, truncated usually 1–2 credits
duplicate_content_checkDuplicate content check url, text – language, mode, url, host, summary, counts, findings, passed, hints, thresholds, pages, limits, note, truncated usually 1–2 credits
hreflang_checkhreflang check url – language, url, counts, findings, pages, matrix, annotations, legend, limits, note usually 1–2 credits
ai_robots_txtAI crawlers in robots.txt site_url, robots_txt, groups, crawlers – file, file_complete, ai_section, crawlers, counts, source, existing, changes, checks, notes, legend, sources, verified_on, language, note usually 1–2 credits
faq_from_supportFAQ from support messages text, faq_url target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv format, input_kind, columns, language, detected_language, languages, counts, selection, faq_page, personal_data, topics, gaps, cited, dropped, limits, placeholder, hints, notes, note, export usually 1–3 credits
price_display_checkPrice display check url, text, html ai: yes | nojurisdiction: auto | de | pl jurisdiction, jurisdiction_auto, jurisdiction_label, source, price, previous_prices, prior_30_days, unit_prices, quantities, discounts, hints, structured, findings, counts, checks, ai, dropped, legal_sources, language, notes, note usually 1–3 credits
form_text_checkForm texts url, html, error_messages target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | svtone: neutral | friendly | formal language, detected_language, tone, required_word, forms, errors, counts, criteria, summary, source, dropped, ai, hints, note usually 1–3 credits
newsletter_checkNewsletter check html, eml, text, subject, preheader target: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv language, detected_language, input, subject, preheader, checklist, counts, links, links_total, images, eml, pairs, improvements, text_cut, dropped, ai, hints, note usually 1–3 credits
privacy_policy_matchPrivacy policy check website_url, pages, policy_url, policy_text – website, pages, policy, not_in_policy, not_found, matched, categories, counts, dropped, ai, limits, legal, notes, check, note, checked_at usually 1–3 credits
social_card_checkSocial card check url – language, url, counts, preview, findings, tags, images, snippet, guidance, limits, note usually 1–2 credits
image_performance_checkImage check url – language, url, counts, total, formats, lcp, page_findings, summary, images, more_images, legend, limits, note, truncated usually 1–2 credits
ai_transparency_checkAI transparency check url, text, html chatbot: auto | yes | nomedia: unknown | no | realistic | artistic | illustrativearticles: unknown | no | unreviewed | reviewedtarget: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv language, detected_language, source, answers, checklist, detected, notices, summary, counts, dropped, ai, legal, hints, note usually 1–3 credits
structured_data_checkStructured data check url, html – language, url, input, counts, formats, findings, items, rules, visible, limits, truncated, note usually 1–2 credits
legal_notice_checkLegal notice check url, text, html ai: yes | nojurisdiction: auto | de | pl jurisdiction, jurisdiction_auto, jurisdiction_basis, jurisdiction_label, source, entity, items, extra, counts, ai, dropped, legal_sources, language, notes, note, truncated usually 1–3 credits
glossary_builderBuild a glossary url, text, html, languages source: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | svmax_pages: 1 | 2 | 3 | 5 | 10 language, language_name, language_source, targets, target_names, items, existing, counts, pages, reading, source, text_cut, dropped, glossary, hints, note, export usually 1–3 credits
local_business_checkBusiness data check url, text ai: no | yes language, url, host, mode, country, table, findings, counts, structured, suggestion, handover, pages, failed, not_reached, limits, ai, dropped, notes, note usually 1–2 credits
consent_preload_checkLoads before consent? url – url, final_url, host, loaded, counts, timing, consent, groups, cookies, storage, notes, check, law, note, truncated, checked_at usually 1–2 credits
ai_literacy_policyAI literacy policy (AI Act Art. 4) text, uses, roles, data, company_name company_size: 1-9 | 10-49 | 50-249 | 250+sector: general | retail | services | crafts | health | finance | education | public_sector | media | it | manufacturing | tourism | socialreview_months: 12 | 6 | 24language: auto | en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv language, title, sections, training, record_columns, markdown, text, profile, counts, dropped, legal, hints, note usually 1–2 credits
withdrawal_button_checkWithdrawal button check url, text, html ai: yes | nojurisdiction: auto | de | pl | eusubscriptions: auto | yes | no jurisdiction, jurisdiction_auto, jurisdiction_basis, jurisdiction_label, law, source, subscription, items, candidates, counts, ai, dropped, legal_sources, language, notes, note, truncated usually 1–3 credits
product_safety_checkProduct page check (GPSR) url, text, html market_language: auto | bg | cs | da | de | el | en | es | fi | fr | hr | hu | it | lt | nl | pl | pt | ro | sk | sl | sv language, market_language, market_language_label, market_basis, page_language, source, product, manufacturer_seat, items, counts, dropped, ai, hints, notes, legal, legal_sources, note usually 1–3 credits
website_checkWebsite check url scope: short | full score, categories, findings, counts, snippet, llms_txt, summary, actions, note fixed 5 credits per task
translateTranslate text target: en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | sv translation, notes, review_status, alignment usually 1–4 credits
proofreadProofreading text text_language: auto | en | de | pl | fr | es | it | pt | nl | cs | sk | uk | ru | bg | ro | hu | sv | da | fi | tr | el | ar | fa | he | hi | zh | ja | ko | th text, html, language, identical, summary, counts, changed_share, changes, notes, beta, note usually 1–4 credits
translate_fileLanguage files text target: en | de | pl | fr | es | it | nl | cs | sk | uk | bg | da | el | fi | hr | hu | lt | pt | ro | sl | svscope: auto | all | missingline_endings: lf | crlf | lf_no_final_newline | crlf_no_final_newline format, target, filename, mime, file, entries, counts, flagged, preview, check, notes, note usually 5–8 credits

Job status

queued and running: still working (position and wait_seconds help you to wait). done: result contains the fields of the tool. failed: nothing was charged (credits is 0); error.code names the reason: address_blocked (Do4U does not open this address, for example a private or internal one; send the text or HTML instead), input_too_large (shorten the input), image_unavailable or job_failed. expired: the result is no longer stored.

Webhooks

Add "callback_url" to POST /api/v1/jobs, and Do4U sends a signed POST (event job.finished with status, credits and the result link - not the result itself) as soon as the job is done. Only public addresses on port 80 or 443; we retry up to 6 times with growing pauses.

Example: create a job with callback_url – and the request Do4U sends to that address when the job is finished:

curl -X POST https://do4u.eu/api/v1/jobs \
  -H "Authorization: Bearer d4u_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"tool": "summarize", "text": "...", "callback_url": "https://example.com/do4u-hook"}'
POST /do4u-hook
Content-Type: application/json
X-Do4U-Event: job.finished
X-Do4U-Signature: t=1793180000,v1=6b1f...

{"event":"job.finished","id":"8f1c0b6a...","tool":"summarize","status":"done","credits":1,
 "created":"2026-10-28T10:00:00+01:00","finished":"2026-10-28T10:00:04+01:00","result_url":"https://do4u.eu/api/v1/jobs/8f1c0b6a..."}

Check the header X-Do4U-Signature: t=<time>,v1=<hex>, where v1 = HMAC-SHA256(secret, "<time>.<raw body>"). Reject old times (e.g. older than 5 minutes). Your secret starts with whsec_ and is shown on the page API keys.

// Node.js: rawBody = the request body exactly as received (string)
const crypto = require('crypto');
function do4uSignatureOk(rawBody, header, secret) {
  const t = (header.match(/t=(\d+)/) || [])[1];
  const v1 = (header.match(/v1=([0-9a-f]{64})/) || [])[1];
  if (!t || !v1 || Math.abs(Date.now() / 1000 - Number(t)) > 300) return false;
  const expected = crypto.createHmac('sha256', secret).update(t + '.' + rawBody).digest('hex');
  return crypto.timingSafeEqual(Buffer.from(expected), Buffer.from(v1));
}
// PHP
function do4u_signature_ok(string $raw, string $header, string $secret): bool {
    if (!preg_match('/t=(\d+),v1=([0-9a-f]{64})/', $header, $m) || abs(time() - (int)$m[1]) > 300) return false;
    return hash_equals(hash_hmac('sha256', $m[1] . '.' . $raw, $secret), $m[2]);
}
$ok = do4u_signature_ok(file_get_contents('php://input'), $_SERVER['HTTP_X_DO4U_SIGNATURE'] ?? '', 'whsec_YOUR_SECRET');

credits.low

When the credit balance of an account drops below the threshold, Do4U sends one signed POST to the webhook address of the account. The event contains no personal data, only the balance, the threshold and the time (UTC). The next one follows only after credits were added again (purchase, monthly free credits or voucher). Signature and secret are the same as for job.finished.

POST /do4u-hook
Content-Type: application/json
X-Do4U-Event: credits.low
X-Do4U-Signature: t=1793180000,v1=6b1f...

{"event":"credits.low","balance":18,"threshold":20,"ts":"2026-10-29T10:00:04Z"}

You set the address and the threshold here: API keys

n8n

Ready-made workflow: a webhook receives the call from Do4U, checks the signature, fetches the result and hands it to your next step.

Download the n8n template (JSON)

Template, not yet tested in a real n8n or Make instance – please test the signature check before you use it.

In n8n: Workflows → Import from file. Then enter your secret in the node "Check signature" and your API key in "Do4U: get result", activate the workflow and use the production URL of the webhook as callback_url. Self-hosted n8n needs the environment variable NODE_FUNCTION_ALLOW_BUILTIN=crypto for the signature check.

Make

  1. Add the module Webhooks → Custom webhook as the trigger and switch on "JSON pass-through" and "Get request headers" in its advanced settings.
  2. Use the address of this webhook as callback_url.
  3. Check the signature: compare v1 from the header X-Do4U-Signature with sha256(t + "." + raw body; hex; your secret).
  4. Add HTTP → Make a request: GET on result_url with the header Authorization: Bearer <your API key>.
  5. Use the field result in your next modules.

Template, not yet tested in a real n8n or Make instance – please test the signature check before you use it.

Recipes: chain tools

A recipe runs 2 to 4 tools one after the other as one job. Send "recipe" instead of "tool": either the name of a ready-made recipe or a list of steps. The text goes into the first step; the result of each step is the text for the next step. GET /api/v1/recipes lists the ready-made recipes, the tools you can use and the field each tool hands on.

curl -X POST https://do4u.eu/api/v1/jobs \
  -H "Authorization: Bearer d4u_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"recipe": [{"tool": "summarize", "options": {"length": "short"}},
               {"tool": "translate", "options": {"target": "pl"}}], "text": "..."}'
curl -X POST https://do4u.eu/api/v1/jobs \
  -H "Authorization: Bearer d4u_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"recipe": "summary_translate_meta", "options": {"target": "pl"}, "text": "..."}'

You get one estimate for all steps (credits_estimate). That amount is reserved at the start. After each step the real use is booked and the rest stays reserved until the end, then it is released. A failed step costs nothing, the remaining steps are skipped, and steps that were already finished are charged.

HTTP/1.1 202 Accepted
{"id": "3c9e0d1a...", "tool": "recipe", "recipe": null, "status": "queued", "model": "do4u-standard", "credits_estimate": 4,
 "steps": [{"n": 1, "tool": "summarize", "credits_estimate": 2}, {"n": 2, "tool": "translate", "credits_estimate": 2}],
 "links": {"self": "https://do4u.eu/api/v1/jobs/3c9e0d1a..."}}

GET /api/v1/jobs/{id} shows every step with its status (pending, queued, running, done, failed, skipped), its credits, a notice input_shortened if the text had to be shortened for the next tool, and the result of each step once the recipe has ended. With no_store the results are shown once. callback_url fires once, at the end of the recipe.

GET https://do4u.eu/api/v1/jobs/3c9e0d1a...

{"id": "3c9e0d1a...", "tool": "recipe", "status": "done", "credits_estimate": 4, "credits": 3, "steps_total": 2,
 "steps": [{"n": 1, "tool": "summarize", "status": "done", "credits": 1, "result": {"summary": "...", "key_points": ["..."]}},
           {"n": 2, "tool": "translate", "status": "done", "credits": 2, "notices": ["input_shortened"], "result": {"translation": "...", "notes": []}}],
 "result": {"translation": "...", "notes": []}}

GET https://do4u.eu/api/v1/recipes

Errors

Errors always have the same form: {"error": {"code": "...", "message": "..."}}. With 429 and 503 the header Retry-After says how many seconds to wait.

CodeHTTPMeaning
missing_key401No API key sent.
invalid_key401Key unknown or revoked.
rate_limited429Too many requests in this minute.
invalid_json400Body is not a JSON object.
unknown_tool400No tool with this name.
invalid_option400Unknown option or value.
invalid_field400A field has the wrong type.
invalid_language400language is not en, de or pl.
empty_input400No text or URL.
text_b_required400compare needs text and text_b.
input_too_long400Input longer than your limit.
input_too_large400Input too large for one job: shorten it or split it into several jobs.
invalid_url400URL not allowed or not readable.
invalid_image400Image missing, wrong format or too large.
payload_too_large413Request body larger than 8 MB.
insufficient_credits402Not enough credits.
email_not_verified402Confirm your email address first.
daily_limit429Free daily limit used up (website_check scope short).
too_many_open_jobs429Too many open jobs in your account.
package_required403The tool needs a purchased package.
busy503The queue is full, try again shortly.
tool_unavailable503The tool is not available right now.
not_found404Unknown endpoint or job.
method_not_allowed405Wrong HTTP method.
internal_error500Error on our side.
invalid_idempotency_key400Idempotency-Key is empty, too long or has characters that are not printable ASCII.
idempotency_conflict422This Idempotency-Key was already used with a different request.
idempotency_in_progress409A request with this Idempotency-Key is still running. Wait a moment (Retry-After).
idempotency_unavailable503Idempotency-Key cannot be used right now. Try again later.
same_language400The text already is in the target language (translate tools). Send allow_same_language: true to translate anyway.

Limits

Credits

A new job shows an estimate (credits_estimate); charged is the actual use (credits). Failed jobs cost nothing. Free credits are used first.

Data

Inputs and results are stored for 7 days (Free) or 30 days (paid packages), then deleted. Images are deleted right after processing. Results are AI-generated: please check them before you use them.

Answer emails without changing your inbox

Do4U does not run an inbox and never stores access data for your mailbox. Two ways to get reply drafts anyway:

S1 – Template for n8n or Make

  1. Trigger: a new email in your mailbox (IMAP trigger in n8n or Make – the access data stays there)
  2. POST /api/v1/jobs with tool support_reply, the email text and no_store: true
  3. Wait a few seconds and fetch GET /api/v1/jobs/{id} until the status is done
  4. Create a draft in your mailbox with the text from result.reply (Gmail or Outlook module)

Download the n8n template (JSON)

In Make, build the same scenario from the modules Email (watch emails), HTTP (make a request), Sleep and Gmail or Outlook (create a draft).

Dev extensions and shop plugins

Extract the ZIP and read the included guide. Load the browser extension locally; check shop plugins in a test shop first. Your own Dev API key is required. Shopware currently provides a CLI.