German Court Decisions API (BGH, BVerfG) for RAG

What does German Court Decisions API do?

German Court Decisions API gives you the official decisions (Urteile und Beschlüsse) of Germany's federal courts as clean, structured JSON, or as ready-to-embed chunks for your vector database. It covers all 84,600 decisions published since January 2010, and it can watch for new ones every day.

Try it now: the prefilled input returns the ten newest BGH decisions in about a minute.

At a glance

	This Actor	Other court decision Actors
Ready-to-embed RAG chunks that stay inside one section	✅	❌
Marginal numbers (Randnummern) kept in every text field, for exact citations	✅	❌
Facts (Tatbestand) and reasons (Gründe) as separate fields	✅	partly
Lower courts split into court, date, file number and type	✅	❌
Daily alerts with a separate position per saved task	✅	partly
Price per 1,000 decisions	$3	$4–10

Based on what the other German court decision Actors in the Apify Store document on their pages (October 2026).

Which courts are covered?

Court	Name	Decisions
BGH	Bundesgerichtshof (Federal Court of Justice)	35,500
BFH	Bundesfinanzhof (Federal Fiscal Court)	11,700
BVerwG	Bundesverwaltungsgericht (Federal Administrative Court)	10,400
BPatG	Bundespatentgericht (Federal Patent Court)	7,300
BAG	Bundesarbeitsgericht (Federal Labour Court)	7,300
BSG	Bundessozialgericht (Federal Social Court)	6,500
BVerfG	Bundesverfassungsgericht (Federal Constitutional Court)	5,800
GmS-OGB	Joint Senate of the federal supreme courts	2

Counts as of October 2026. The data comes from rechtsprechung-im-internet.de, the official case law portal of the Federal Ministry of Justice.

Why use this Actor?

• Built for RAG and AI agents. In chunk mode every chunk stays inside one section (Leitsatz, Tenor, Tatbestand, Gründe ...), breaks between paragraphs or at real sentence ends, never ends on a heading, and carries the court, file number, title, heading path and marginal numbers (Randnummern). Your AI can cite "BGH, IX ZB 72/08, Rn. 4–6" instead of guessing.
• Daily case law alerts. Turn on Only new or updated since the last run and schedule the Actor. You get only what the portal published since the last run, and you only pay for that. Every saved task keeps its own position, so a tenancy-law task and an employment-law task never get in each other's way.
• You pay per decision, not per chunk. One decision with 40 chunks costs the same as one decision without chunks. Decisions removed by your filters are free.
• Precise filters. Court, decision date, decision type (Urteil, Beschluss ...), file number (Aktenzeichen) and keywords in the full text and cited norms, matching any or all keywords. Umlauts are handled: "Kuendigung" finds "Kündigung".
• Clean data. No HTML. ECLI, cited norms, headnote (Leitsatz), operative part (Tenor), facts (Tatbestand) and reasons (Gründe) as separate fields, plus the full text as Markdown with headings and marginal numbers. Lower courts come split into court, date, file number and decision type, so you can follow a case through the instances. Tables become one line per row.
• Reliable. Interrupted downloads resume where they stopped. If the portal refuses the direct connection, the Actor switches to Apify Proxy by itself, for the rest of the run.

	This Actor	The official portal
Bulk download with filters	✅	❌ one decision at a time
Full text as clean JSON and Markdown	✅	❌ XML inside ZIP files
Facts and reasons as separate fields, with marginal numbers	✅	❌
Lower courts split into court, date and file number	✅	❌
Section-aware RAG chunks with marginal numbers	✅	❌
Only new decisions since the last run	✅	❌
Schedules, webhooks, API, MCP, Make, Zapier, n8n	✅	❌

How to get German court decisions

1. Choose the courts, a date range, decision types, keywords or file numbers. Leave everything empty to get the newest decisions of all courts.
2. Pick the output format: one item per decision, or RAG chunks.
3. Click Start and download the results as JSON, CSV or Excel, or read them through the API.

How to get alerts for new case law

1. Set your filters, for example court BAG and keyword Kündigungsschutz.
2. Turn on Only new or updated since the last run and set Updated since to yesterday, so the first run starts from today.
3. Save the input as a task and add a daily schedule.
4. Add a webhook, an email or a Slack integration to the task to be told when new decisions arrive.

The portal publishes new decisions every working day, usually 20 to 90 across all courts. Large batches in which the portal re-stamps thousands of old decisions at once (as in June 2025) are skipped, so you are not charged for old case law.

How much does it cost to get German court decisions?

This Actor uses pay-per-event pricing. You pay a small fee per run and a fee per decision, and nothing for the platform resources the run uses.

Event	Price
Decision returned	$0.003 ($3 per 1,000 decisions)
Run started	$0.005 (Apify counts it once per GB of memory; the default is 1 GB)

Examples:

• 100 BGH decisions as RAG chunks: about $0.31.
• A daily alert in one field of law that finds three new decisions a day: about $0.42 per month.
• All 84,600 decisions once: about $254.

You can cap every run with Maximum decisions or the run's spending limit. The Actor stops cleanly at the limit and never charges for a decision it did not deliver.

Input

All fields are optional. Example: BGH decisions on tenancy law (Eigenbedarf) since 2025, as RAG chunks.

{
  "courts": ["BGH"],
  "dateFrom": "2025-01-01",
  "keywords": ["§ 573 BGB", "Eigenbedarf"],
  "outputFormat": "chunks",
  "chunkSize": 2000,
  "maxDecisions": 200
}

Field	What it does
courts	Courts to include. Empty means all.
dateFrom, dateTo	Decision date range.
decisionTypes	For example Urteil or Beschluss.
keywords, keywordMode	Words or phrases in the full text or cited norms; any or all.
fileNumbers	Specific file numbers, e.g. IX ZB 72/08.
outputFormat	decisions (one item per decision) or chunks.
includeFullText	Add facts, reasons and the full decision as Markdown. Turn off for a light citation index.
chunkSize, chunkOverlap	Target chunk length in characters, and paragraphs repeated between chunks.
maxDecisions	Stop after this many decisions. Newest first; in monitoring mode the oldest pending change first, so nothing is skipped.
onlyNew, updatedSince	Monitoring: only decisions added or changed since the last run or since a date.

Output

Chunk mode

{
  "docId": "KVRE467122601",
  "ecli": "ECLI:DE:BVerfG:2026:es20261014.2bve000526",
  "court": "BVerfG",
  "courtName": "Bundesverfassungsgericht",
  "chamber": "2. Senat",
  "decisionDate": "2026-10-14",
  "fileNumber": "2 BvE 5/26",
  "decisionType": "Einstweilige Anordnung",
  "title": "Erfolgreicher Eilantrag im Organstreitverfahren - Verpflichtung einer Bundesministerin zur Depublikation eines im Instagram-Profil ihres Ministeriums veröffentlichten Videoclips - ...",
  "norms": ["Art 21 Abs 1 S 1 GG", "§ 32 Abs 1 BVerfGG", "§ 63 BVerfGG"],
  "url": "https://www.rechtsprechung-im-internet.de/jportal/?quelle=jlink&docid=jb-KVRE467122601&psml=bsjrsprod.psml&max=true",
  "section": "gruende",
  "sectionLabel": "Gründe",
  "headingPath": ["A.", "II."],
  "marginalNumberFrom": 11,
  "marginalNumberTo": 15,
  "text": "[11] b) Sämtliche in dem Video thematisierten Themenkomplexe berührten unmittelbar das Aufgabengebiet der Antragsgegnerin ... \n\n### B.\n\n[15] Der Antrag auf Erlass einer einstweiligen Anordnung hat Erfolg.",
  "chunkIndex": 5,
  "chunkCount": 20
}

Decision mode (default)

One item per decision. facts, reasons, dissentingOpinion and fullTextMarkdown are left out when Include full text is off. Long texts are shortened here:

{
  "docId": "STRE202650160",
  "ecli": "ECLI:DE:BFH:2026:U.180626.IIIR38.23.0",
  "court": "BFH",
  "courtName": "Bundesfinanzhof",
  "chamber": "3. Senat",
  "decisionDate": "2026-06-18",
  "fileNumber": "III R 38/23",
  "decisionType": "Urteil",
  "title": "(Zur Gewinngrenze für die Inanspruchnahme des Investitionsabzugsbetrags nach § 7g EStG)",
  "norms": ["§ 7g EStG 2009", "§ 7g Abs 1 S 2 Nr 1 EStG 2009", "§ 2 Abs 2 S 1 Nr 1 EStG 2009", "EStG VZ 2020"],
  "lowerCourts": [
    {
      "relation": "vorgehend",
      "court": "Finanzgericht Baden-Württemberg",
      "date": "2023-05-02",
      "fileNumber": "10 K 1873/22",
      "decisionType": "Urteil",
      "text": "vorgehend Finanzgericht Baden-Württemberg, 2. Mai 2023, Az: 10 K 1873/22, Urteil"
    }
  ],
  "headnote": "NV: Gewinn im Sinne von § 7g Abs. 1 Satz 2 Nr. 1 des Einkommensteuergesetzes (EStG) ist der steuerliche Gewinn gemäß § 2 Abs. 2 Satz 1 Nr. 1 EStG ...",
  "orientationHeadnote": null,
  "operativePart": "Auf die Revision des Beklagten wird das Urteil des Finanzgerichts Baden-Württemberg vom 02.05.2023 - 10 K 1873/22 aufgehoben.\n\nDie Klage wird ...",
  "url": "https://www.rechtsprechung-im-internet.de/jportal/?quelle=jlink&docid=jb-STRE202650160&psml=bsjrsprod.psml&max=true",
  "sourceUpdatedAt": "2026-10-14T21:08:55.134Z",
  "facts": "I.\n\n[1] Streitig ist, ob die Voraussetzungen für die Bildung eines Investitionsabzugsbetrags (IAB) nach § 7g Abs. 1 des Einkommensteuergesetzes (EStG) ...",
  "reasons": "II.\n\n[12] Die Revision ist begründet. Das FG hat zu Unrecht die Voraussetzungen für die Inanspruchnahme des IAB als erfüllt angesehen. ...",
  "dissentingOpinion": null,
  "fullTextMarkdown": "# BFH 3. Senat, Urteil vom 2026-06-18, III R 38/23\n\n## Leitsatz\n\nNV: Gewinn im Sinne von § 7g Abs. 1 Satz 2 Nr. 1 ..."
}

facts, reasons and dissentingOpinion keep the marginal numbers ([12] ...), so every passage can be cited. facts is set when the decision has a separate Tatbestand (mostly judgments); otherwise the whole reasoning is in reasons. Some short orders of the BGH and BVerfG carry their brief reasons inside the operative part; the portal publishes them that way, so for those reasons is null and the text is in operativePart. If the parties waived the reasons (§ 313a ZPO, common for BAG judgments), facts and reasons are null and the note about the waiver is only in fullTextMarkdown. A lower-court line in an unusual form keeps only text, with the other fields null. The Output tab has a table view for each mode.

Use the API or an AI agent

Python:

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("layerwiselabs/german-court-decisions").call(
    run_input={"courts": ["BAG"], "dateFrom": "2026-01-01", "outputFormat": "chunks"}
)
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["fileNumber"], item["marginalNumberFrom"], item["text"][:80])

AI agents (Claude, Cursor, VS Code ...) can call the Actor through the Apify MCP server by adding layerwiselabs/german-court-decisions as a tool. LangChain and LlamaIndex can load the dataset through their Apify integrations.

Who is it for?

Who	What for
Legal tech and AI startups	Retrieval for legal assistants, with citable marginal numbers
Law firms and in-house legal teams	Daily alerts on new case law in their fields
Tax advisors	New BFH decisions every week
Patent attorneys	BPatG and BGH patent decisions
Researchers	Datasets for German legal NLP and empirical legal studies

FAQ

Is this data available for free elsewhere?

Yes. German court decisions are official works without copyright (§ 5 UrhG), and the federal government publishes them for free on rechtsprechung-im-internet.de and in the trial phase of its new portal rechtsinformationen.bund.de. This Actor saves you the work around it: downloading and parsing the XML, cleaning the text, splitting it for RAG, keeping track of what is new, and delivering it to your tools on a schedule.

Which courts are not covered?

State courts (Amtsgerichte, Landgerichte, Oberlandesgerichte, state administrative and labour courts) are not published on this portal and are not included. The portal starts with decisions from January 2010.

How fresh is the data?

The portal publishes new decisions every working day, usually a few days after the decision date. Every run reads the current table of contents, so it sees everything published up to that moment.

How do keywords work?

A keyword matches anywhere in the full text or the cited norms, ignoring case, extra spaces and umlaut spelling. Phrases work as typed, for example § 573 BGB. With Keyword match set to All keywords, a decision must contain every keyword.

Can I export to CSV or Excel?

Yes. Every run's dataset downloads as JSON, CSV, Excel, XML or HTML, and the Output tab has table views for decisions and for chunks.

Does it handle personal data?

The courts anonymize decisions before publication. The Actor returns the published text unchanged and adds no personal data.

What happens if the portal is slow or down?

Downloads resume where they broke off and are retried with increasing pauses. A decision that cannot be downloaded is skipped for that run and, in monitoring mode, retried in the next run. Only if it still fails in three runs in a row, while other decisions download fine, is it left out for good. If no decision can be downloaded at all, the run fails instead of returning an empty result.

Something is missing or wrong?

Open an issue on the Issues tab with the file number and what you expected.