LeXML-mcp
MCP server for LexML Brasil — search the legislative SRU acervo, resolve URNs, and inspect service metadata.
README
LeXML-mcp
MCP server for LexML Brasil — search the legislative SRU acervo, resolve URNs, and inspect service metadata. Built with FastMCP + httpx.
Features
| Tool | Description |
|---|---|
lexml_search |
Search the LexML SRU acervo with CQL queries |
lexml_resolve_urn |
Resolve a LexML URN to its public URL |
lexml_explain |
Retrieve metadata about the SRU service |
Every response includes a provenance envelope (source_url, source_kind, authority, retrieved_at, content_hash, backend_name, backend_version, schema_version).
Quick start
# Install
uv sync
# Run the server (stdio transport)
uv run lexml-mcp
The server listens on stdio — configure your MCP client to launch it:
{
"mcpServers": {
"lexml": {
"command": "uv",
"args": ["run", "--directory", "/path/to/LeXML-mcp", "lexml-mcp"]
}
}
}
Usage examples
lexml_search
# Search for documents about consumer protection law
result = await lexml_search(
query='dc.title any "código de defesa do consumidor"',
start_record=1,
maximum_records=10,
record_schema="dc"
)
Returns structured records with Dublin Core fields, number_of_records, next_record_position, and provenance.
lexml_resolve_urn
# Resolve a URN to its public URL
result = await lexml_resolve_urn(
urn="urn:lex:br:federal:lei:1990-09-11;8078"
)
Returns public_url, status_code, redirect_chain, and provenance.
lexml_explain
# Inspect SRU service capabilities
result = await lexml_explain()
Returns SRU explain record data with server info, database info, and supported schemas.
Challenge detection
The SRU endpoint may return an HTML verification page instead of XML. The server detects this and returns a structured object:
{
"is_challenge_like": true,
"content_type": "text/html",
"raw_excerpt": "<!DOCTYPE html>...",
"status_code": 200,
"provenance": { ... }
}
Error handling
All errors are returned as structured objects — never crashes:
| Error type | Trigger |
|---|---|
network_error |
Connection refused, DNS failure |
timeout |
Request exceeded timeout (default 15s) |
invalid_xml |
SRU returned malformed XML |
http_error |
Non-2xx HTTP status |
Caching
LRU cache with configurable TTL (default 300s, max 256 entries). Challenges are never cached as successes. Configure via environment:
LEXML_CACHE_TTL— seconds (default300)LEXML_CACHE_MAXSIZE— max entries (default256)
Configuration
| Env variable | Default | Description |
|---|---|---|
LEXML_SRU_URL |
https://www.lexml.gov.br/busca/SRU |
SRU endpoint |
LEXML_URN_TEMPLATE |
https://www.lexml.gov.br/urn/{urn} |
URN resolution template |
LEXML_TIMEOUT |
15.0 |
HTTP request timeout (seconds) |
LEXML_CACHE_TTL |
300 |
Cache TTL (seconds) |
LEXML_CACHE_MAXSIZE |
256 |
Max cache entries |
Project structure
src/lexml_mcp/
├── server.py # FastMCP server, tool registration
├── config.py # Environment-based configuration
├── connectors/
│ └── acervo.py # SRU search, explain, URN resolution
├── models/
│ ├── provenance.py # Provenance envelope
│ ├── search.py # Search result models
│ └── error.py # Structured error objects
└── utils/
├── cache.py # LRU TTL cache
└── challenge.py # Challenge HTML detection
Testing
uv run pytest
29 tests covering:
- SRU success parsing (Dublin Core records)
- Challenge HTML detection
- Malformed XML handling
- Timeout simulation
- URN resolution
- Explain response parsing
- Provenance envelope construction
- Cache set/get/expiry/eviction
- Structured error types
- Tool registration
Limitations
- Parser, linker, renderer not implemented — these require JVM/Scala/Haskell infrastructure and are roadmap items
- SRU endpoint may return challenge HTML — the server detects and reports this transparently
- No multi-tenancy — single-user, local-first design
- No DOCX output — renderer integration is future scope
- No Unicode normalization — preserves original content as-is
- No external health checks — liveness is local only
Dependencies
mcp— official Python SDK (FastMCP)httpx— async HTTP client- stdlib only beyond those two
License
GPL-3.0-or-later
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。