Free vs Paid Search APIs for AI Agents the original metered category.
Search was billed per query long before AI agents existed, because every search costs the provider index and crawl infrastructure. With 1,012 live servers, this is Loomal's biggest category — and the one where per-call pricing is most native.
Search is the largest category in Loomal's index — 1,012 live servers — and the one with the longest history of usage-based billing. Maintaining a fresh web index is brutally expensive, which is why commercial web search has been metered per query for two decades. Agents didn't change that economics; they just became the new caller.
The free side of the category is real too, but it's a different kind of search: local and domain-specific, where the index is small or already public.
Search MCP servers on the Loomal Index
Metabase
Lets AI clients search, explore, query, and visualize data in a Metabase instance.
mcp-server-search
MCP server for web search operations
Serena MCP: the IDE for your agent
A powerful toolkit for coding, providing semantic retrieval and editing capabilities.
Airweave Search
MCP server for searching Airweave collections with natural language queries.
Claude Code Explorer MCP
Explore the Claude Code CLI source — browse tools, commands, search code, and more.
exa
Fast, intelligent web search and web crawling. New mcp tool: Exa-code is a context tool for coding
Tolgee
Your app's translations in Tolgee: search keys, create translations, trigger machine translation
socraticode
MCP server for local codebase indexing, semantic search, and code dependency graphs.
arxiv-mcp-server
Search arXiv papers, download full text, semantic search, citation graphs, and alerts via MCP.
brightdata-mcp
Bright Data's Web MCP server enabling AI agents to search, extract & navigate the web
tavily-mcp
MCP server for advanced web search using Tavily
AutoTS
Automated time series forecasting with model search, anomaly detection, and event risk analysis
Showing 12 of 1012 live Search servers — browse them all on the marketplace.
Where free search genuinely works
Free, open-source search servers thrive where no expensive index is needed. socraticode and Serena do semantic retrieval over your own codebase — the corpus is local, so there's nothing to meter. arxiv-mcp-server searches arXiv, a public archive with a free API, adding full-text download and citation graphs on top. Metabase lets agents query and visualize data in a Metabase instance you already run.
The pattern: when the index is yours or publicly funded, free self-hosted search is the right default. Your only costs are compute and the rate limits of public upstreams.
Why web-scale search is almost never free
General web search is different. Servers like exa, tavily-mcp, and brightdata-mcp front providers that crawl and index the live web — infrastructure measured in data centers, not Docker containers. Those upstream providers bill by usage (check each provider's docs for current tiers), and any free quota is a marketing allowance, not an economic model.
For agents this matters doubly, because agents search constantly. A research agent can issue hundreds of queries per task, blowing through free tiers built around human search behavior in minutes.
Why x402 fits search better than any other category
Search is the canonical per-call product: every query has a marginal cost to serve and a clear unit of value to the caller. x402 lets that unit be priced directly — the agent sends a query, receives an HTTP 402 with the price, pays in USDC, and gets results, settled on Base in roughly two seconds with a signed receipt. Minimum price is $0.01 per call; no API key signup, no monthly tier to size in advance.
Compare that to subscription tiers, where you pre-purchase a quota and either waste the remainder or hit the ceiling mid-task. For unpredictable agent workloads, paying exactly per query is the honest pricing model.
Choosing a search server
Match the index to the job. Searching your own code or data: free local servers, no contest. Searching public archives like arXiv: free servers over free APIs, within their rate limits. Searching the live web: expect to pay someone, and prefer pricing that scales with your actual query count.
Loomal's Search category lists all 1,012 live servers with descriptions and x402 per-call pricing where maintainers have configured it — the fastest way to see which web search options an agent can pay for autonomously.
Frequently asked questions
Should my agent use a free or paid search MCP server?
Use free servers when the corpus is local or public — codebase search with Serena or socraticode, paper search via arxiv-mcp-server. For live web search, the index is expensive to maintain, so plan on metered access and pick pricing that matches your query volume.
How does pay-per-call compare to a subscription for search APIs?
Search subscriptions sell pre-sized quotas, which agents either exhaust mid-task or waste. x402 pay-per-call charges per query in USDC from $0.01, settled in about two seconds, so cost tracks usage exactly — better for bursty or unpredictable agent workloads.
Why do agents hit search rate limits so fast?
Free tiers and rate limits are calibrated to humans typing one query at a time. A single research agent can issue hundreds of queries per task, so it consumes a day's human quota in minutes. Per-call payment removes the quota problem: each query simply pays its own way.
Where can I compare search MCP server options?
Loomal's Search category — the largest in the index at 1,012 live servers — shows each listing's description, package type, and x402 pricing where configured, putting free local tools and paid web search endpoints side by side.
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