Docs

Validate OpenRTB over MCP

A coding agent working on bidder or exchange code can validate the payloads it writes, as it writes them. The Model Context Protocol server exposes the full rule set as tools.

Hosted: zero install

{
  "mcpServers": {
    "rtblint": { "url": "https://rtblint.org/mcp" }
  }
}

The hosted endpoint speaks MCP over HTTP and exposes validate_bid_request, validate_bid_response (with an optional bid_request argument for cross-validation), and list_openrtb_versions.

Local: stdio

cargo install rtblint-mcp

{
  "mcpServers": {
    "rtblint": { "command": "rtblint-mcp" }
  }
}

Same tools, running entirely on your machine, for environments where payloads must not leave the network.

Why give an agent a validator

Agents writing OpenRTB integration code confidently produce payloads with moved fields and stale enums; the training data is full of 2.5-era examples. A validator tool turns that guesswork into a checked loop: the agent validates against the exact spec snapshot you target and fixes findings by rule id before you ever review the code. Full endpoint details are in the MCP reference.