Lawbot
Lawbot is a pure offline desktop software for legal document desensitization. Officially, it claims to complete sensitive information identification, pseudonymization and Markdown export of Word, PDF and image files on the local machine, enabling legal documents to safely enter the AI workflow.
Updated: 2026-10-26
What is Lawbot?
The official positioning of Lawbot is a "pre-processing tool for AI collaboration on legal documents", with the slogan "Desensitize first, then enter the AI workflow". It addresses a very specific dilemma: lawyers want to hand contracts, pleadings and judgments to AI, but are worried about disclosing clients' information together with them. The official solution is a pure offline desktop software. The product architecture has no internet module from the very beginning, and the official even states that "it can still be used after unplugging the network cable". Compared with online desensitization services, the difference is that it shifts the risk from "whether the service provider is trustworthy" to "the data never leaves the local machine".
Who is it for?
- Law firm teams and in-house legal departments handling non-litigation contracts, transaction documents and legal opinions.
- Litigation teams that need to submit judgments, pleadings, and evidence materials to external AI for processing.
- Legal professionals whose primary materials are scanned documents or image evidence and who are concerned about data leakage.
- A handler who has multiple documents under one project and needs unified pseudonymization rules.
Key features
6Pure Offline Local Processing
The official states that the product architecture has no internet access module from the very beginning. It will actively intercept all external network requests. The recognition model and OCR model are downloaded with the package, and all processing is done locally on the computer.
Sensitive Information Identification and Review
The official claims it covers 18 categories of Chinese sensitive information, with the model achieving an identification accuracy of 95% and continuous improvement. It completes manual review through hit coloring and one-click supplementary entry.
Pseudonymization and Masking
The official offers pseudonymization or masking processing. The same subject maintains the same pseudonym, which is suitable for contracts and project materials that need to retain the semantic structure.
Structured Markdown Export
The official product is said to export clean, lightweight Markdown, which can be used for summarization, translation, review, risk list creation, and draft generation.
Authorization Backfilling
The official description states that the results processed by AI can have pseudonyms backfilled under authorized conditions, forming a closed loop of desensitization, AI analysis, and restoration.
Memory Workbench
The official states that the pseudonymization rules are unified across files, and .tmm memory files are used to save review progress, pseudonymization rules, and whitelists/blacklists, allowing users to resume work with a double-click.
How to use Lawbot
- The official minimum configuration is Windows 10 (64-bit) or above, or macOS, with 8GB or more of RAM and 2GB or more of available hard disk space.
- for scanned documents or batch processing, 16GB of RAM and an SSD are recommended.
- Scanned PDFs require page-by-page OCR, and the official notes that low-configured machines may take as long as tens of minutes.
- you can first use other tools to convert them into editable text and then process them with Tuomincat.
- The official states that offline processing is designed to build a privacy barrier, but once the exported Markdown is handed over to an external AI, it is still bound by the other party's terms, so it is important to distinguish between the two types of risks.
- Pricing and authorization are differentiated by personal version and enterprise version.
- The official states that the personal version starts at 499 yuan, and the commercial version is subject to the official page before use.
- The official states that the recognition accuracy is 95% and continuously improving, which means there is still a possibility of missed detection, and manual review must be done before delivery.
Traffic & popularity
Data month Sep 2026
Lawbot receives about 297.5K visits per month.
Monthly visits
10 monthsEstimated from third-party traffic data.
