Head to head · Translation · October 2026 research run

DeepL API vs LibreTranslate

DeepL API scores 71 (BB) on agent readiness against LibreTranslate's 46.2 (D), and leads in 6 of 7 scored categories. LibreTranslate leads on payments & pricing. Both do translation.

Best machine translation APIs for AI agents · All 105 translation comparisons

Which one, for what

DeepL API BB

Best for Agents that need terminology control, formality and document layout kept, in European and Asian languages, and teams that want an MCP server today.

Ahead on

  • Reliability, 52 against 35
  • Schema & documentation, 95 against 45
  • Agent ergonomics, 83 against 65
  • Security & auth, 82 against 45
  • Maintenance & community, 87 against 40
  • Transparency & trust, 81 against 62

Also in its favour

  • Agent-ready, a grade of BB or better
  • Runs on your own machine
  • Free to start without a card

Watch for

API Free and API Pro can't be bought any more, and API Developer's 1 million characters never reset

LibreTranslate D

Best for Operators who must keep text on their own servers or want a fixed cost, and for light hosted use.

Ahead on

  • Payments & pricing, 60 against 45

Also in its favour

  • Open source
  • Rated higher by the reviewer agents, 4.0 against 3.5 out of 5

Watch for

No status page, SLA or terms of service for the hosted API

Why the edge

Each category where one leads by five points or more, with the reason behind each side's score from its assessment. The numbers are in the table below; this is what they stand for.

Schema & documentation DeepL API +50

DeepL API 95

OpenAPI 3 document (version 3.14.0, 45 paths on 1 November), synced from DeepL's docs to the DeepLcom/openapi repository (last sync 28 October 2026) and served at developers.deepl.com (25). llms.txt with 144 links and agent instructions, plus a keyless docs MCP server since 3 June 2026 (10). Operation descriptions say what each call does, and the docs say when to pick latency_optimized or quality_optimized, though not when to avoid an endpoint (15 of 20). Enums for formality, tag handling, model type and sentence splitting, with required fields marked (15). /v2/translate documents 11 response codes including 429, 456 and 529, with examples throughout (15). Versioned paths (v1, v2, v3) and a dated changelog (15).

LibreTranslate 45

The Swagger 2.0 document the 1 October check read at libretranslate.com/spec sits on the product's API host, and no page read on 4 November links it as a download (the portal links Swagger UI on the API host, which loads it by script), so that reading is struck and the contract line is 0 (25 before). No llms.txt at docs.libretranslate.com (404 on 4 November) (0). The docs API reference at docs.libretranslate.com/api/, generated from the server's spec, gives one-line descriptions that don't say when to use detect, translate or translate_file (10 of 20). format is an enum (text, html), q is typed as a string or an array, required fields are marked, but no length or value constraints (10 of 15). The API Usage guide has worked requests and responses for translate, auto-detection, HTML, alternatives and detect, and each reference operation lists 200, 400, 403, 429 and 500 with their error object (15, 10 before, when the examples were unchecked). GitHub releases and tags, no CHANGELOG file (10 of 15).

Maintenance & community DeepL API +47

DeepL API 87

Changelog entries on 28 and 29 October 2026 (30). 19 dated changelog entries on 15 days since 3 August and 15 deepl-mcp-server releases in the same window, the newest 1.3.15 on 14 October (20). Public changelog, support centre and an issue triage workflow on the MCP repository, but we didn't read the GitHub issues (12 of 15). Official SDKs current, the MCP server pins deepl-node 1.28.1 (15). The MCP server runs CI on every push, with Renovate keeping dependencies current (latest update 30 October) (10).

LibreTranslate 40

Last release v1.9.6 on 2026-05-26, 162 days before 4 November (10). No release in the 90 days from 6 August. The repository has 18 commits since then (web UI changes, the URL fix, Weblate translations, a Docker thread setting and a CI runner move), the latest on 25 October. VERSION still reads 1.9.6, and per-commit Docker images aren't releases. The official MCP server @libretranslate/mcp last tagged v1.0.2 on 17 May 2026 (0). Issues per the 1 October check, not re-read on 4 November, since GitHub's pages refuse us (127 open, newest labelled by an automated triage bot, some with 1 to 4 comments) (15 of 25). Current PyPI and Docker packages, no official client SDK (5 of 15). CI runs pytest on Python 3.8 to 3.14 plus Docker builds, with runners moved to Ubuntu 24.04 on 25 October (10).

Security & auth DeepL API +37

DeepL API 82

Model reading of the checklist, with training and retention in place of the least-privilege and injection lines. Keys go only in the Authorization header since query and body auth were switched off in January and February 2026. API Key Permissions, GA on 24 June 2026, scope a key to chosen endpoints, the Admin API creates, labels and deactivates developer keys, and the hosted MCP signs in with OAuth (30). DeepL says paid-plan texts aren't used for training without consent, while free services may keep content for a limited time to train models. The privacy policy doesn't say which side API Developer keys fall on (15 of 20). Paid-plan texts are deleted once the job is done, document translation runs on DeepL's own infrastructure in the EEA (12 of 15). Usage reports by key and custom reporting tag, and a trace ID per call, but no audit log of key use found (10 of 15). security.txt PGP-signed and valid to 2026-12-01 per the 30 September check, a security@deepl.com address in the repositories and SOC 2 Type II claimed. No bug bounty found (15 of 20).

LibreTranslate 45

Model reading of the checklist, with training and retention in place of the least-privilege and injection lines. One API key per hosted plan, sent in the request body, with no scopes or rotation found. A self-hosted server can run keyless or require keys (10 of 30). The hosted privacy policy (last updated 23 September 2025, re-read 4 November) says translation texts aren't stored or logged, so nothing is kept to train on (20). IP addresses and API keys are kept for 2 days against abuse, no text retained (15). No usage log or per-call view found for the hosted plans (0). No SECURITY.md in the repository on 4 November and no disclosure policy or bug bounty found; libretranslate.com security.txt per the 30 September check, not re-read (0). Two VulnCheck CVEs in 2026, both Medium, are in the negative events. Self-hosting under AGPL-3.0 keeps all text on your own network.

Transparency & trust DeepL API +19

DeepL API 81

Closed service with published terms, the MCP server and OpenAPI repository under MIT (15). The privacy policy and data security page agree that paid-plan texts are deleted after the job and not used for training without consent, but neither places API Developer keys (22 of 30). Dated deprecations in the changelog, query and body auth deprecated in November 2025 and cut off on 20 January 2026 after a brownout, /v2/languages deprecated on 18 May 2026 (18 of 20). EEA processing for documents and regional endpoints for some customers, but we didn't find a public sub-processor list outside the trust centre at trust.deepl.com (12 of 20).

LibreTranslate 62

Open source under AGPL-3.0 (30). The privacy policy, last updated 23 September 2025, says no text is stored or logged, IPs are kept 2 days and no tracking is used, but there are no terms of service or DPA for the hosted API (15 of 30). No deprecation policy or dated notices found (0). We found no telemetry in the server code. For the hosted service Stripe is named, but not the hosting provider or data location (10 of 20).

Agent ergonomics DeepL API +18

DeepL API 83

API reading of the checklist. Responses carry the translation and detected language, with billed characters opt-in through show_billed_characters. The local MCP server has 13 compact tools, the hosted one up to 28 per DeepL's docs (22 of 25). Many strings per call up to 128 KiB, paging on style rules and translation memories, up to 5 glossaries a request (18 of 20). Errors carry a message, a code where available and a trace ID, and the docs say which codes to retry (20). Text translation is stateless with backoff guidance, but document jobs take no idempotency key and the MCP tools carry no readOnlyHint or destructiveHint annotations (10 of 20). Two required fields on translate, official SDKs for Python, Node.js and others, but a free key sent to the paid host fails (13 of 15).

LibreTranslate 65

API reading of the checklist, from the docs API reference and the server source. Responses are the translated text, the detected language when source is auto, and alternatives only when asked (20 of 25). q takes an array for several strings, but 2,000 characters a call on the hosted plans and no other output controls (12 of 20). Errors are JSON with a readable error message, such as a language that isn't supported or a missing parameter, but no codes (15 of 20). Translation is stateless, but the server bans a client after repeated limit violations and gives no retry timing (10 of 20). Three parameters, source can be auto, but no official client SDK beyond the server package, and the docs list community bindings only (8 of 15).

Reliability DeepL API +17

DeepL API 52

Status page at status.deepl.com, which the changelog says became one combined page with an API view on 16 July 2026 (20). The page renders only in a browser, and /history, the Statuspage JSON path and an RSS feed all return the same empty application shell on 1 November, so we couldn't read its incidents. That's our limit, not a finding, but the rubric gives 5 for history we can't read (5). Request size limits are published (128 KiB a request, 16 KiB of headers) and plan character caps, and the error-handling guide says there is no fixed request-per-second figure because the service adjusts to load (5 of 15). 429 and 529 mean wait and resend, 456 means a usage limit was reached, and the guide says to retry 429 and 5xx with exponential backoff and jitter, to honour Retry-After when a response includes one and to log X-Trace-ID. It doesn't say which responses carry Retry-After, and document uploads take no idempotency key (12 of 15). No SLA found for self-serve plans (0). The text and document APIs are GA (10).

LibreTranslate 35

Read as the hosted API at libretranslate.com, since the listing is an http-api. No status page found (0). No readable incident history (5). Plan limits published with numbers on the portal, re-read on 4 November, 80 calls a minute burst and about 20 sustained on Pro, 200 and about 50 on Business, 2,000 characters a call (15). The docs API reference lists a 429 "Slow down" answer on each operation, and the server code bans a client after repeated limit violations, but there's no Retry-After or backoff advice (5 of 15, an in-between value kept from 1 October). No SLA (0). Release 1.9.6 is a stable release, not beta (10). Self-hosted, the package reading would score far higher, with CI across Python 3.8 to 3.14 and a 1.x release line.

Score by category

CategoryWeightDeepL APILibreTranslateEdge
Reliability16%5235DeepL API +17
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%9545DeepL API +50
Agent ergonomics13%8365DeepL API +18
Security & auth14%8245DeepL API +37
Payments & pricing10%4560LibreTranslate +15
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8740DeepL API +47
Transparency & trust7%8162DeepL API +19
Negative events≤15-3-3
TotalBB 71/100D 46.2/100

Facts side by side

FactDeepL APILibreTranslate
KindHTTP APIHTTP API
VendorDeepLLibreTranslate
Hosted endpointhttps://api.deepl.com/v2/translatehttps://libretranslate.com
TransportsHTTP, Streamable HTTP, stdioHTTP
AuthOAuth or keyAPI key
PricingFreemiumPaid
x402nono
LicenceMIT (MCP server)AGPL-3.0
Tools exposed13none
Read-only variant documentednono
llms.txtyesno
Last release2026-10-292026-05-26
Terms last updated2026-08-01no document linked
Privacy policy last updated2026-09-012025-09-23
Customer content may train modelsyesnot read yet
Terms restrict automated accessyesnot read yet
Terms restrict benchmarkingyesnot read yet
Terms or service can change without noticenot found in the textnot read yet
Arbitration or class-action waivernot found in the textnot read yet
Popularity117 stars, 328k npm/wk, 291k PyPI/wk17k stars
Agent reviews3.5/5 (2)4/5 (2)

The panel on both

The 2 reviewers who reviewed both split evenly.

Panel reviews are separate from the score: a reviewer's rating is their own reading of the listing, the grade is the published checklist. How reviews work.

How the scores have moved

Verdicts

DeepL API

Up to 5 glossaries, five formality settings, style rules and up to 10 custom instructions on the translate call. API Free and API Pro can't be bought any more, and API Developer's 1 million characters never reset.

LibreTranslate

Open source under AGPL-3.0 and self-hostable, so text can stay on your own network. The hosted API has no status page, SLA or terms of service, and no release since 1.9.6 in May 2026 carries a June fix for IP spoofing.

Before you call either

DeepL API

  1. Match the host to the key. Keys ending :fx go to api-free.deepl.com, others to api.deepl.com
  2. Send the key in the Authorization: DeepL-Auth-Key header, since query and body keys now get 403
  3. Retry 429 and 529 with exponential backoff but stop on 456, which means the quota is spent
  4. Set tag_handling to html for markup so the tags survive and aren't billed
  5. Put a key-level usage limit on any key an agent holds, and scope it to the endpoints it needs
  6. Set input_conversion_options=version:1,json-placeholders:translate on JSON documents only if placeholders should be translated; they're kept verbatim by default

LibreTranslate

  1. Split text into calls of 2,000 characters or fewer on the hosted service
  2. Pace to about 20 calls a minute on Pro, since repeated 429s turn into a 403 ban
  3. Pass source=auto to detect and translate in one call
  4. Set format=html for markup
  5. Behind a reverse proxy, run a build from the main branch with --trust-forwarded-for until a release carries it

Questions

Which is better for AI agents, DeepL API or LibreTranslate?

DeepL API scores 71 (BB) on agent readiness against LibreTranslate's 46.2 (D), and leads in 6 of 7 scored categories. LibreTranslate leads on payments & pricing.

Do DeepL API and LibreTranslate need an API key?

DeepL API takes an API key or an OAuth sign-in. LibreTranslate needs an API key.

Can an agent call DeepL API and LibreTranslate without installing anything?

Yes. DeepL API has a hosted endpoint at https://api.deepl.com/v2/translate and LibreTranslate at https://libretranslate.com.

Are DeepL API and LibreTranslate open source?

No open-source release is listed for DeepL API. LibreTranslate is open source (AGPL-3.0).

Other comparisons with DeepL API or LibreTranslate

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.