# Did It Work?

> Did It Work? is a free MCP server for charity and giving. Read-only charity evidence for AI agents: cited evaluator figures and caveats. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Free
- **Platforms:** API
- **Website:** https://diditwork.org/agents/mcp
- **MCP Registry name:** org.diditwork/did-it-work
- **Categories:** charity and giving
- **Listing updated:** 2026-11-15

## About

Did it work? gives assistants a careful way to answer donor questions. It does not score or rank charities; it passes on what independent evaluators published, with the evaluator, period and source attached to each figure and a note that figures are modeled, not guaranteed. The docs list 22 read-only tools, no key and no stored input.

The suggested first call is decision_brief with the donor's question in their own words. It returns the sentence to say first, a shortlist of up to five charities in the site's own order, an evaluator-fund alternative, three questions to ask and a ready-to-quote paragraph. verdict_chain walks one charity from L0 Defined to L7 Checked, marks links it does not cover and names the weakest one without adding them up. After the person decides, report_decision_effect records changed, confirmed or not_deciding as an anonymous count.

It fails closed on empty input, unknown arguments, instructions hidden in a question or unknown charities. The endpoint is stateless Streamable HTTP and supports several protocol versions. Setup is shown for Claude, ChatGPT, Cursor and VS Code.

## Features

- **Decision brief:** One call returns an answer, shortlist, questions to ask and a quotable paragraph.
- **Verdict chain:** Link-by-link evidence view for one charity with the weakest link named.
- **Cited figures:** Each number carries evaluator, period and source.
- **Outcome tally:** Anonymous count of whether answers changed decisions.
- **No score or ranking:** Never adds its own grade across causes.
- **Fails closed:** Rejects empty, unknown or instruction-bearing input.

## Good for

- Answer where to give for a stated cause
- Check the evidence behind a named charity
- Respond to a claim that a charity's results were debunked
- Compare an evaluator fund with picking one charity
- Quote evaluator figures with proper caveats

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http did-it-work https://api.diditwork.org/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://api.diditwork.org/mcp
```

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

```json
{
  "mcpServers": {
    "did-it-work": {
      "url": "https://api.diditwork.org/mcp"
    }
  }
}
```

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

```json
{
  "servers": {
    "did-it-work": {
      "type": "http",
      "url": "https://api.diditwork.org/mcp"
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "did-it-work": {
      "serverUrl": "https://api.diditwork.org/mcp"
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "did-it-work": {
      "type": "http",
      "url": "https://api.diditwork.org/mcp"
    }
  }
}
```

## Tools

- `decision_brief`: One-call answer to a donor question with shortlist and quotable text
- `verdict_chain`: Link-by-link evidence chain for one charity
- `report_decision_effect`: Record whether the answer changed, confirmed or did not affect the decision
- `search_orgs`: Search charities with fixed answers
- `get_claim`: Fetch one cited claim by ID
- `estimate_impact`: Say-as impact sentences for a named charity and budget

## FAQ

### Do I need an API key?

No. It is public, read-only and needs no account.

### Does it rate charities?

No. It declines scoring requests and gives the evaluators' own labels.

### Does it store my question?

The docs say nothing you send is stored.

### Which clients are covered?

Claude, ChatGPT, Cursor and VS Code, plus curl.

### Is Did It Work? free?

Yes. Did It Work? is free to use.

### How do I add Did It Work? to Claude Code?

Run this command in your terminal: claude mcp add --transport http did-it-work https://api.diditwork.org/mcp

---
Source: https://ai-powered.com/mcp/did-it-work
