# GroundLens

> GroundLens is an open-source MCP server for ai models and llm tooling. Deterministic checks of LLM answers against sources, with PASS, REVIEW or FAIL. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, Linux, macOS, Windows
- **Website:** https://groundlens.dev
- **Repository:** https://github.com/groundlens-dev/groundlens
- **MCP Registry name:** io.github.groundlens-dev/groundlens
- **Categories:** AI models and LLM tooling, developer tools
- **Listing updated:** 2026-11-15

## About

GroundLens is built around figures. Before anyone acts on an LLM answer, it checks that every number appears in the retrieved passages or follows from them by a sum, difference, ratio or percentage change, compared as exact fractions in the locale and units the documents use. It also checks that a figure belongs to the right company, contract or year, and that the answer matches what was asked, such as percent versus euros.

The decision is PASS, REVIEW or FAIL under your policy, and the signed, hash-chained record is stored before the decision returns. The site says no model, network, account or API key is involved, and that an optional LLM judge only sees answers the exact checks could not reject. It is explicit about limits: a real figure from the wrong row of the right table, or a fact wrong in the passages themselves, is not caught.

The MCP server exposes three tools, and there are Langfuse, OpenTelemetry and Grafana integrations. It installs from PyPI and is Apache-2.0.

## Features

- **Exact figure checks:** Verify numbers and derived arithmetic against passages.
- **Scope check:** Fail real figures taken from another company, contract or year.
- **Intent check:** Flag answers in the wrong kind or unit of figure.
- **Signed records:** Hash and store a record before returning a decision.
- **Judge escalation:** A FAIL never reaches the LLM judge, which can only escalate.
- **Observability hooks:** Send decisions to Langfuse or OpenTelemetry.

## Good for

- Gate financial or legal answers before they reach a user
- Audit past answers with offline record verification
- Add a deterministic pre-check ahead of an LLM judge
- Trace guardrail decisions in Langfuse or Grafana

## Install

### Claude Code (Terminal)

```sh
claude mcp add groundlens -- uvx groundlens
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "groundlens": {
      "command": "uvx",
      "args": [
        "groundlens"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "groundlens": {
      "command": "uvx",
      "args": [
        "groundlens"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "groundlens": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "groundlens"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "groundlens": {
      "command": "uvx",
      "args": [
        "groundlens"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "groundlens": {
      "command": "uvx",
      "args": [
        "groundlens"
      ]
    }
  }
}
```

## Tools

- `verify_answer`: Check an answer against its source passages.
- `verify_run`: Verify a run of checks.
- `verify_records`: Verify stored signed records.

## FAQ

### Does it need a model or network?

No. The site says it runs in your process with no server, account, API key, model or network.

### What does it return?

PASS, REVIEW or FAIL with reasons, and a signed record.

### What does it not catch?

Wrong rows of the right table, sentences the passages never say, and errors in the passages themselves.

### Is it free?

Yes, Apache-2.0.

### Is GroundLens free?

Yes. GroundLens is open source, so you can use and self-host it for free.

### How do I add GroundLens to Claude Code?

Run this command in your terminal: claude mcp add groundlens -- uvx groundlens

### What are the best alternatives to GroundLens?

The closest alternatives to GroundLens are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Ouroboros](https://ai-powered.com/mcp/ouroboros): Ouroboros is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Concord MCP](https://ai-powered.com/mcp/concord-mcp): Concord MCP is a freemium MCP server for developer tools.

---
Source: https://ai-powered.com/mcp/groundlens
