# esg-mcp-servers [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/freminder/esg-mcp-servers  
**GitHub Stars:** 0  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/esg-mcp-servers

## Description
ESG metric extraction, PDF processing, vector search, and regulation analysis

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "esg-mcp-servers": {
    "command": "uvx",
    "args": ["esg-mcp-servers"],
    "env": {
      "POSTGRES_DSN": "",
      "MONGODB_URI": "",
      "ANTHROPIC_API_KEY": "",
      "EMBEDDING_MODEL": "",
      "EMBEDDING_DIMENSIONS": ""
    }
  }
}
```

**Requires environment variables:** `POSTGRES_DSN`, `MONGODB_URI`, `ANTHROPIC_API_KEY`, `EMBEDDING_MODEL`, `EMBEDDING_DIMENSIONS` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What esg-mcp-servers does

The esg-mcp-servers MCP server groups 31 tools into six server processes for ESG document workflows. The metrics extractor handles ESRS-aligned data across emissions, energy, water, waste, social, and governance topics. It also provides free-text question answering, semantic keyword search, query-domain detection, emissions query classification, and batch KPI extraction with database persistence.

The PDF processor validates reports, extracts text and tables, supports OCR fallback for tables, creates embeddings, and can run the full processing pipeline. The vector store manages document chunks and query-response caching in PostgreSQL with pgvector. The regulation server downloads configured EU regulations from EUR-Lex, parses and indexes them, and searches their articles semantically.

The scraper searches for ESG reports, crawls company websites for PDF links, downloads files into MongoDB GridFS, and exposes job status and cancellation operations.

## How it works

Install the Python package, provide the database and model configuration, start the supporting services, and run the migration command. Documents can be discovered or downloaded, processed into chunks and embeddings, stored in PostgreSQL, and queried through semantic search or RAG tools. Regulation files follow a similar download, extraction, embedding, and storage flow.

Each capability can run as its own MCP process: `esg-metrics-extractor`, `esg-pdf-processor`, `esg-vector-store`, `esg-regulations`, or `esg-scraper`. `esg-mcp-all` exposes all 31 tools from one process. This layout lets a client load only the server groups it needs.

## Setup and configuration

Install with `pip install esg-mcp-servers`. PostgreSQL 16 with pgvector is required for vector storage, metrics, and regulations. MongoDB 7 is required for PDF binary storage. Anthropic access is required for RAG queries and LLM-based metric extraction. The repository includes Docker Compose support for starting local databases, followed by `esg-mcp-migrate` for schema setup.

Set `POSTGRES_DSN`, `MONGODB_URI`, and `ANTHROPIC_API_KEY` in the environment. `EMBEDDING_MODEL` defaults to `Snowflake/snowflake-arctic-embed-l-v2.0`, while `EMBEDDING_DIMENSIONS` defaults to `1024`. Install the optional `[scraper]` extra when using Selenium-dependent scraping for JavaScript-heavy sites.

Claude Desktop can register each executable separately or use the combined `esg-mcp-all` command. The project is licensed under MIT.

## Tools and capabilities

The esg-mcp-servers MCP server supports these workflow areas:

- Extract ESRS-aligned environmental, social, and governance metrics.
- Validate, parse, chunk, and embed PDF reports.
- Store document vectors and cache query responses.
- Download and search EU regulation articles.
- Discover ESG reports and manage scraping jobs.
- Run all listed tools through one combined MCP process.

## Limitations and notes

The README identifies PostgreSQL, MongoDB, and Anthropic requirements by feature rather than as a single universal dependency set. The scraper’s Selenium dependency is optional but needed for JavaScript-heavy sites. The provided client configuration documents Claude Desktop; other clients are not described in the project material. Embedding model and vector dimensions must remain compatible with the stored embeddings and database configuration.

_Full upstream README: https://allmcps.com/mcp/esg-mcp-servers/readme_