Skip to main content
Glama

Get started

DPOLens runs on your own servers with Docker Compose, so your policies stay there. Get started takes you from the first command to your first question, and Connect your AI assistant sets up Claude Code, Codex, Cursor and GitHub Copilot in VS Code.

Related MCP server: legal-mcp

What's inside

  • An MCP server with four tools to search and read the law and your policies, quoting each clause word for word with its key, its version and the days it applies, including a duty that starts later. The MCP server

  • Laws as packs, each from its official source: GDPR and the UAE PDPL today, and more as their packs are released. Supported laws

  • Your own policies, from Word or Markdown files, read into clauses without an AI model and published as versions that never change. Your own policies

  • Search by words and by meaning, in English and Arabic, on the CPU, with no LLM and no API key. Questions written beside each clause help it order what it found, and are never shown. Search questions

  • A dashboard and two logs: a governance log that is append-only and hash-chained, and every question asked, with personal data taken out. The governance log

  • An HTTP API, the same one the dashboard and the MCP server use. The HTTP API

Measured

  • Search: on held-out questions, the clause the guidance points at is among the top five answers 73% of the time for GDPR, and 53% for the UAE's law. How it was measured

  • Use: Claude Code with the plugin searched before writing code in 36 of 36 runs of tasks that touch personal data, and in none of the 45 runs of tasks that do not. How it was measured

Contributing

Law packs are the contribution that matters most, and CONTRIBUTING.md says what one needs. A security problem is reported privately, as SECURITY.md says.

Licence

The code is Apache-2.0, with DCO sign-off and no CLA. packs/ carries its own licence: the law text keeps its source's, and the structure around it is CC-BY-4.0.

Available Tools

4 tools
get_clauseRead one clauseA
Read-only
Inspect

Read one clause by its key, with where it sits and what sits beneath it.

Use it when a search result points at another clause, or when you need the whole of an article rather than the part that matched. The call is recorded, with the key that was read rather than a question.

Inputs: key (required), lang, as_of, whole_subtree. Returns: One clause, where it sits, and what sits beneath it.

get_documentRead one document's outlineA
Read-only
Inspect

Read one document and its top level structure, without the text.

The outline is where to find a clause key when you know roughly where to look. Read the words with get_clause. The call is recorded.

Inputs: slug (required), as_of. Returns: One document and its top level. The text is left out on purpose: a document of…

list_documentsList what can be citedA
Read-only
Inspect

List the laws this community version can cite, and which version of each is in force.

Inputs: as_of, limit, offset. Returns: What this instance can cite today.

search_policiesSearch the lawA
Read-only
Inspect

Search the privacy law this community version holds.

Call this before writing or changing code that collects, stores, logs, shares or deletes personal data, and before answering a question about what the law requires. What comes back is the clause itself, word for word, with the key, document and version to cite it by.

Questions asked here are kept 30 days, with nothing about who asked.

Inputs: query (required), lang, laws, as_of, +2 more. Returns: What a search found, best match first.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.1/5.0

    Scored across 4 tools

    Disambiguation4/5

    Each tool has a distinct retrieval shape: query-based search, key-based clause read, document listing, and document outline read. The only near-pair is get_clause vs. get_document, but the descriptions explicitly disambiguate by stating one returns words and the other returns structure without text.

    Naming Consistency5/5

    All four tools follow a clean verb_noun snake_case pattern (search_policies, get_clause, list_documents, get_document). The verbs match the operation type precisely, with no mixed conventions.

    Tool Count5/5

    Four tools is well within a sensible range and each maps to a necessary step in a read-only legal-lookup workflow: discover documents, browse outline, search, and read a clause. Nothing feels padded or missing for the scope.

    Completeness4/5

    The read-only surface covers the core loop of search → locate → read clause, plus document/version discovery. Minor gaps remain: no way to request a specific historical version of a clause, and get_document only exposes top-level structure, forcing extra get_clause calls for deep article traversal.

    Maintenance

    ActivityActive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      D
      maintenance
      Provides AI-powered access to the largest EU GDPR enforcement decisions database, enabling semantic search, GDPR article lookup, and enforcement statistics across all EU/EEA Data Protection Authorities.
      4
      21 PyPI
      1
      MIT
    • A
      license
      Not graded
      quality
      D
      maintenance
      A remotely callable MCP server for US legal research that provides tools to search, retrieve, and analyze US case law from the CourtListener API, enabling agents to build evidence packs from primary sources without generating legal content.
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Verifies legal citations by fetching the exact, in-force wording of cited articles from official PL and EU legal sources, preventing hallucination by returning verbatim text.
      2
      7
      Apache 2.0