Skip to main content
Glama
wabarakat

agent-marketplace-mcp

by wabarakat
README.md
# agent-marketplace-mcp

<!-- mcp-name: io.github.wabarakat/agent-marketplace-mcp -->

MCP server for the [Agent Data Marketplace](https://dlfyresearch.com), a data
marketplace built for AI agents. Agents search data sellers (finance, satellite,
agriculture, weather, geo, research and more) and request datasets. No API key.

## Install

**Hosted (no install):** add this URL to any MCP client that supports remote servers:
```
https://dlfyresearch.com/mcp
```
Claude Code: `claude mcp add --transport http agent-marketplace https://dlfyresearch.com/mcp`

**Local (stdio)** via this package:

Claude Code:
```bash
claude mcp add agent-marketplace -- uvx agent-marketplace-mcp
```

Claude Desktop, Cursor and other clients (`mcpServers` JSON):
```json
{
  "agent-marketplace": {
    "command": "uvx",
    "args": ["agent-marketplace-mcp"]
  }
}
```

## Tools
| Tool | What it does |
|---|---|
| `list_categories` | Data categories with data point counts |
| `search_datapoints` | Search by keywords, category, seller, max price, verified providers |
| `get_datapoint` | One data point |
| `search_sellers` | Search sellers |
| `get_seller` | Seller profile and its data points |
| `request_data` | Purchase, sample, or describe data that isn't listed |
| `get_request_status` | Status of a request |

## Pricing
Prices are in credits: 1 credit = $0.01 USD. Each data point also shows `price_usd`.
Credits can't be purchased yet and nothing is charged: requests are queued and
you can check their status with `get_request_status`.

## Configuration
| Variable | Default |
|---|---|
| `AGENT_MARKETPLACE_URL` | `https://dlfyresearch.com` |

Requests from this server send `User-Agent: agent-marketplace-mcp/<version>`.

The HTTP API is also usable directly: see https://dlfyresearch.com/llms.txt and
https://dlfyresearch.com/openapi.json.

## License
MIT

TDQS

A3.6/5.0

Scored across 7 tools

Disambiguation5/5

Each tool targets a distinct resource and action: listing categories, searching data points, searching sellers, fetching individual details, requesting data, and checking request status. The search vs. get pairs are clearly differentiated by keyword-based discovery versus id-based retrieval, and the two search tools operate on different entity types.

Naming Consistency5/5

All tools use a consistent snake_case verb_noun convention (list_categories, get_seller, search_datapoints, get_datapoint, search_sellers, request_data, get_request_status). The only slight variation is the three-part get_request_status, which remains predictable and readable.

Tool Count5/5

Seven tools is well-scoped for a data marketplace: category listing, seller discovery, data point discovery, detail views, and the request/status lifecycle. Each tool has a clear role and no redundant tools appear present.

Completeness4/5

The consumer-side marketplace workflow is well covered from browsing categories and sellers through requesting data and checking status. Minor gaps remain around managing existing requests, such as listing all requests or cancelling a queued request, but agents can work around these.