
Taiwan legal data, built for people and for AI
Every Taiwanese court judgment, searchable and citable.
Dr.Legal (法律偵探) is a free legal platform built and run by a practicing Taiwanese lawyer. It covers 22.8 million court judgments, every statute article, constitutional rulings and administrative rules, and exposes the same corpus to AI assistants through TLR, an open MCP server.
Counts are a dated snapshot (2026-11-18), not a live query. Sources: Judicial Yuan and Taiwanese government open data.
What you can use
TLR: Taiwan Legal RAG for AI assistants
A retrieval-only MCP server and REST API. Your AI tool calls it to search judgments, read full text, look up statute articles and administrative rulings, and every result carries a verifiable citation. No generation on our side.
Connection guideJudgment search
Full-text and structured search over the whole corpus: case number, statute, keyword, Boolean and semantic modes, with the original judgment rendered on a public page you can cite.
Open search (Chinese interface)AI legal Q&A
Ask in plain language; the system retrieves relevant judgments first and answers with citations. Generation runs on an overseas cloud model provider through our own legal reasoning harness; it is not an on-premises system.
Open Ask (Chinese interface)Enterprise on-premises deployment
For organizations whose documents cannot leave their own network: legal LLM and retrieval deployed in your own server room, with a quantified hallucination audit and ISO 42001 governance alignment.
Enterprise servicesHow it is built, honestly
Retrieval first, then answers
The AI never answers from memory alone. It searches the corpus first and cites the judgments it actually read; if nothing relevant is found, it says so instead of inventing a case.
Retrieval on our own servers, generation in the cloud
The judgment database and retrieval system run on our own servers. AI answers are generated by a large language model from an overseas cloud model provider, driven by a legal reasoning harness we designed and tuned over more than a thousand hours. The online service is not an on-premises system; do not enter real names, national ID numbers or confidential case details.
Self-governed to ISO 42001
Risk assessment, model monitoring and anomaly detection follow the ISO/IEC 42001 AI management system standard. We have not applied for third-party certification and say so plainly.
Source-available
The TLR retrieval tool is published on GitHub under the Elastic License 2.0 and is the most-starred Taiwanese legal repository there.
Partnerships and licensing
Data feeds, API access for your product, or on-premises deployment: write to us and we will reply personally.
[email protected]