Loomal

How to Monetize Search MCP Servers with x402

Per-query pricing built the commercial search API market. x402 brings the same model to MCP — without API keys, quotas, or a billing department.

Search is the largest category in the Loomal index — 1,012 live servers — and the one where per-call monetization needs the least explaining, because the search industry already runs on it. Commercial web search APIs have charged per thousand queries for years. What x402 changes is the plumbing: instead of API keys, prepaid credit packs, and monthly invoices, the agent pays for each query inline, in USDC, settled on Base in about two seconds before your handler runs.

The category's range is wide. exa and tavily-mcp front intelligent web search; brightdata-mcp adds search, extraction, and navigation; arxiv-mcp-server does semantic search over academic papers with citation graphs; socraticode indexes codebases for semantic retrieval. Different corpora, same billable unit: the answered query.

Search already proved the per-query model — agents just broke the old billing

Every serious search provider meters by the query, because cost (index lookups, ranking compute) and value (one answered question) both arrive one query at a time. What agents broke is everything wrapped around that meter. An autonomous agent can't fill in a signup form, store an API key, or pick the right monthly tier for a workload it doesn't know yet. It can read an HTTP 402 response, pay the stated price from its wallet, and retry — which is the entire x402 flow.

For a search server operator this means the addressable market stops being 'developers who registered for my API key' and becomes 'any agent that discovers my endpoint.' Loomal's index is where that discovery happens, across the 1,012 search servers currently live.

Pricing logic: queries are cheap, answers are not

The floor works hard in this category. At Loomal's $0.01 minimum, a hundred queries cost an agent a dollar — trivially affordable, yet enough to sustain an operator whose marginal query cost is a fraction of that. Plain keyword or index search belongs at or near that floor.

Move up as the query does more. Semantic search over a curated corpus — the arxiv-mcp-server or socraticode pattern, where you maintain embeddings and citation or dependency graphs — supports $0.02 to $0.05 per query. Search that bundles retrieval with extraction or navigation, as brightdata-mcp does, carries real proxy and rendering costs and prices accordingly. If your tool returns full documents rather than result lists, consider splitting search and fetch into separately priced tools so agents pay for exactly what they take.

From listing to revenue on Loomal

Search servers in the index already have marketplace pages; claiming yours means verifying GitHub ownership of the repo behind it. Then connect a remote endpoint — Loomal probes it and publishes the live tool list, so an agent developer comparing dozens of search servers sees your actual query interface — and set your per-call price in the console. Minimum $0.01; repricing later is a single field edit.

Each paid call settles in USDC on Base with no chargeback exposure, and arrives with an Ed25519-signed receipt. The platform fee is 5% of settled volume, currently waived.

Standing out in a thousand-server category

Search is the most crowded category on the index, and generic web search is where the crowding concentrates. The monetization edge goes to specialized corpora: a translation-key search like Tolgee's domain, a natural-language layer over business data the way Metabase's server exposes dashboards, an academic or code-specific index. Specialized search faces less price pressure, attracts higher-intent agent traffic, and is harder to substitute — three properties that matter a lot more when every call has a price on it.

Frequently asked questions

Why would an agent pay for search when free search engines exist?

Free engines serve humans pages of links; agents need structured, rate-limit-free, programmatic results. Agents already consume paid search APIs heavily today — x402 just removes the API-key and subscription friction. A specialized index the agent can't get elsewhere makes the choice even simpler.

What should one search query cost?

Start at Loomal's $0.01 minimum for straightforward queries and move up with the work per call: $0.02–$0.05 for semantic search over a maintained corpus, more when a query bundles extraction or full-document retrieval. Pricing search and fetch as separate tools keeps the cheap path cheap.

How do agents find my search server among a thousand others?

Through the Loomal index, where listings expose live-probed tool lists, categories, and pricing. Claimed listings with a connected endpoint and accurate tool schemas rank ahead of unclaimed ones in practice, because agent developers can verify what they're integrating before the first paid call.

Do I need to manage API keys or rate limits for paying agents?

No keys — payment is the credential. Every call either carries a settled x402 payment or gets a 402 response; there's no account state to manage. You may still want rate limiting for abuse protection, but you no longer need it as a billing enforcement mechanism.

Run a Search MCP server?

Claim your listing, set a per-call USDC price, and let AI agents pay for every call over x402.

List it on Loomal