In a move that screams 'do one thing and do it cheaply,' Plainserp has launched a new API specifically designed to feed clean Google search results directly to AI agents. Priced at a staggering $0.30 per 1,000 queries, the service strips away the visual clutter of modern search engines, returning only titles, links, and snippets as raw JSON. This isn't just another scraping wrapper; it’s a deliberate attempt to lower the token overhead for agents that are currently choking on knowledge panels and ads.

The Anti-Bloat Philosophy

The core pitch here is token efficiency. Plainserp argues that when an LLM reads search results, it doesn't need the AI Overviews, the 'People Also Ask' boxes, or the shopping carousels. Those elements are just noise that the model has to pay to process and then ignore. By returning only organic results in Google's order, with position numbers and thumbnails where available, Plainserp claims to give agents exactly what they need to write their own overviews without the bloat. The service explicitly excludes AI Overviews, knowledge panels, and ads, positioning itself as a utility for builders who want raw data, not a polished consumer experience.

Technical Specs and Limitations

Under the hood, the API is straightforward. It supports parameters for query, number of results (1-20), pages (1-6), country, language, and time filters. You can pull up to 120 results per query by requesting multiple pages. Latency is reported between one to two and a half seconds, which is acceptable for most agent workflows but won't win any speed contests. The service is independent and not affiliated with Google, meaning it fetches public results and parses them. This independence comes with a caveat: if Google changes its HTML structure, Plainserp might go down. However, the company promises you won't be charged for failed requests, a nice touch for developers worried about budget overruns during a scraping breakage.

Integration and Pricing Model

Integration is built for the modern agent stack. Plainserp offers an MCP (Model Context Protocol) server at plainserp.com/mcp, allowing any compatible client to instantly gain a web_search tool. There’s also a skill file for coding agents to easily add search functionality to projects. The pricing model is pure pay-as-you-go with no monthly plans or minimums. You can start with a $6 top-up, and each search costs $0.0003. The first 1,000 searches are free, requiring no credit card. Default rate limits are set at 300 searches per minute, with options to request higher limits via the dashboard.

Key Takeaways

  • Plainserp offers Google search results as clean JSON for $0.30 per 1,000 queries, focusing on token efficiency for AI agents.
  • The service explicitly excludes AI Overviews, ads, and knowledge panels to reduce noise and cost for LLMs.
  • Integration is supported via a dedicated MCP server and skill files for coding agents.
  • Pricing is pay-as-you-go with no monthly fees, starting at $6, and includes 1,000 free trial searches.
  • As an independent scraper, it faces potential outages if Google changes its SERP structure, though failed requests aren't charged.

The Bottom Line

Plainserp is a necessary evil for the current state of agent economics, trading stability for radical cost savings. If you can tolerate the risk of a scraping breakage, this is the most efficient way to ground your agents in real-time data without burning your API budget on useless UI elements.