# Agent evidence

> Agent evidence is an open-source MCP server for ai models and llm tooling. Offline MCP tools that check an AI claim against evidence you supply. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI
- **Repository:** https://github.com/probityai/probity/tree/HEAD/packages/agent-evidence-mcp
- **MCP Registry name:** io.github.probityai/agent-evidence-mcp
- **Categories:** [AI models and LLM tooling](https://ai-powered.com/categories/ai-and-llm), [developer tools](https://ai-powered.com/categories/developer-tools)
- **Listing updated:** 2026-11-18

## About

The server returns one of three decisions for a claim: supported, contradicted or not_established. The bundled sample shows the idea: if a passage quoted in a report is missing from the source capture, verify_sample with the contradicted scenario returns the reason selected_span_missing_from_record.

It belongs to the agent-evidence family, an open record format, with probityai doing the checking and grading. Nine tools are read-only, offline and bounded, and nothing is fetched, written or sent. Besides claim verification they replay conformance vectors, explain a vector, look up vocabulary terms, check artifact digests, canonicalize JSON and verify DSSE envelopes.

Version 0.2.5 installs from PyPI or runs through uvx, and there is a Claude Code plugin. The audience is agent builders, auditors and platform engineers. It is Apache-2.0 licensed.

## Features

- **Claim verification:** Decides whether a claim holds against evidence bytes you provide.
- **Offline and read-only:** Nothing is fetched, written or sent.
- **Conformance vectors:** Lists corpora, runs vectors and explains individual ones.
- **Digest and DSSE checks:** Checks artifact digests and verifies DSSE envelopes.
- **JSON canonicalization:** Provides a canonicalize_json tool.

## Good for

- Verify an AI report's quoted passages against the source capture
- Run conformance vectors for an evidence format
- Check an artifact digest in a pipeline
- Audit whether an agent's claim is supported

## Install

### Claude Code (Terminal)

```sh
claude mcp add agent-evidence-mcp -- uvx agent-evidence-mcp
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "agent-evidence-mcp": {
      "command": "uvx",
      "args": [
        "agent-evidence-mcp"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "agent-evidence-mcp": {
      "command": "uvx",
      "args": [
        "agent-evidence-mcp"
      ]
    }
  }
}
```

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

```json
{
  "servers": {
    "agent-evidence-mcp": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "agent-evidence-mcp"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "agent-evidence-mcp": {
      "command": "uvx",
      "args": [
        "agent-evidence-mcp"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "agent-evidence-mcp": {
      "command": "uvx",
      "args": [
        "agent-evidence-mcp"
      ]
    }
  }
}
```

## Tools

- `verify_claim`: Decides whether a claim holds against supplied evidence.
- `verify_sample`: Runs a built-in sample claim with a chosen scenario.
- `list_vector_corpora`: Lists the shipped conformance vector corpora.
- `run_vectors`: Replays conformance vectors.
- `explain_vector`: Explains a single vector.
- `lookup_term`: Looks up a vocabulary term.
- `check_artifact_digest`: Checks an artifact digest.
- `canonicalize_json`: Canonicalizes JSON.
- `verify_dsse`: Verifies a DSSE envelope.

## FAQ

### What decisions can it return?

Supported, contradicted or not_established.

### Does it touch the network?

No. It is offline and read-only.

### How do I install it?

pip install agent-evidence-mcp==0.2.5, or run it with uvx in Claude Code.

### What licence?

Apache-2.0.

### Is Agent evidence free?

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

### How do I add Agent evidence to Claude Code?

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

### What are the best alternatives to Agent evidence?

The closest alternatives to Agent evidence are REA, Context7, Chrome DevTools MCP, Codebase Memory and GitHub.

## Where this comes from

- Description: written from the project's README.
- Install steps: from the official MCP Registry entry.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Context7](https://ai-powered.com/mcp/context7): Context7 is a freemium MCP server for developer tools.
- [Chrome DevTools MCP](https://ai-powered.com/mcp/chrome-devtools-mcp-2): Chrome DevTools MCP is an open-source MCP server for browser automation.
- [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.
- [awesome-mcp-servers](https://ai-powered.com/mcp/awesome-mcp-servers-2): awesome-mcp-servers 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.

---
Source: https://ai-powered.com/mcp/agent-evidence
