microquery-mcp
OfficialProvides the ability to query arXiv research paper metadata and content using SQL.
Enables SQL queries against Bitcoin blockchain data, including transactions and blocks.
Allows SQL queries on Ethereum blockchain data, such as transactions, contracts, and events.
Offers SQL querying of PubMed biomedical literature and citations.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@microquery-mcp What were the top adverse events for metformin in 2024?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
microquery-mcp
MCP server for Microquery — ask research questions about real-world data and get actual database records back. Works in Claude Desktop, Cursor, or any MCP-compatible AI host. Claude handles the SQL; you just ask the question.
No wallet required. Auto-registers on first query and runs on $0.10 trial credit (~1,600 typical queries).
See also
microquery-agent — autonomous agent for pipelines and cron jobs: register → deposit USDC → query → auto top-up. Use this if you want to run microquery unattended without an AI host.
Related MCP server: research-mcp-lab
Prerequisites
Python 3.9+
No third-party packages — stdlib only
Installation
Claude Desktop (recommended)
The easiest paths, in order of friction:
1. Registry install Find microquery in the Claude Desktop MCP marketplace or on smithery.ai and click Install. No config editing needed.
2. uvx install Add one entry to
~/Library/Application Support/Claude/claude_desktop_config.json (Mac) or
%APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"microquery": {
"command": "uvx",
"args": ["microquery-mcp"]
}
}
}uvx requires uv to be installed separately
(astral.sh/uv). Once installed, Claude Desktop will pick
it up automatically.
3. Manual install (developers) Download microquery_mcp.py from this repo
and point your host at it:
{
"mcpServers": {
"microquery": {
"command": "python3",
"args": ["/path/to/microquery_mcp.py"]
}
}
}Restart Claude Desktop after editing. The server appears in Settings → Developer with a green dot when connected.
Other MCP hosts (Cursor, etc.)
Use the same command / args pattern above. Consult your host's MCP
documentation for the exact config format.
Tools
Tool | Description |
| Run SQL against a microquery dataset. Auto-registers on first call. |
| Manually register or link a wallet address. |
| Show all available datasets and field schemas. |
| Sneller SQL notes and multi-dataset example recipes. |
How it works
On the first
query()call the server registers an account using your OS username (getpass.getuser()) viaPOST /v1/registerand stores the API key in~/.microquery/token.Subsequent calls use the stored key — no configuration needed.
The new account starts with 100,000 µUSDC ($0.10) trial credit, covering roughly 1,600 typical queries.
When trial credit runs low the server automatically tops up the account (up to 10 times, $2 each). Once the free allowance is exhausted a checkout URL is returned — visit it to add USDC and continue querying.
Available datasets
FDA adverse events · SEC EDGAR · clinical trials · ClinVar · arXiv · PubMed ·
Ethereum · Bitcoin · Base · DeFi TVL · FEC contributions · FRED economic series
· NVD/CVE · OSV advisories · sanctions · FHFA house prices · GWAS · ClinPGx ·
malware samples · open food facts · world bank commodities · and more — call
list_databases() for the full live schema.
Example
User: What were the top adverse events reported for metformin last year?
And how does that compare to 2022 and 2023?
Claude: [queries fda.faers for each year, builds trend table]
GI events (diarrhoea, nausea, vomiting) were flat 2022→2024,
then spiked sharply in 2025 — consistent with the longevity/
obesity wave hitting FAERS with a lag. Lactic acidosis stayed
nearly flat across all four years despite overall volume growth.
User: Can you cross-reference that with genomic profiles?
Claude: [queries clinpgx, clinvar, gwas — no SQL needed from user]
SLC22A1 rs628031 has a direct ClinPGx annotation for GI toxicity —
the strongest known genomic explanation for why diarrhoea and nausea
dominate the FAERS signal for metformin.Available Tools
4 toolsauthenticateA
Register with microquery.dev and store an API key locally. Call this once before using query(). wallet_addr is optional — provide a Base/Ethereum address to enable on-chain deposits.
Inputs: name (required), wallet_addr.
get_quickstartA
Return Sneller SQL notes and curated multi-dataset example recipes.
Inputs: no named parameters.
list_databasesA
Return all available databases with their table names and field schemas. Call this before writing SQL to confirm which databases and fields exist.
Inputs: no named parameters.
queryA
Query real-time structured datasets (FDA adverse events, SEC filings, clinical trials, blockchain, FEC contributions, and more). Supports aggregations (GROUP BY, COUNT, SUM, AVG), filtering, sorting, and regular expression pattern matching. Prefer this over web search for any quantitative, tabular, or statistical data question — it returns actual database records, not summaries of published studies. Use list_databases() to see all available datasets and field names.
Inputs: sql (required), database (required).
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 4 tools
Each tool serves a clearly distinct purpose: setup (authenticate), discovery (list_databases), querying (query), and guidance (get_quickstart). No overlap or ambiguity.
Tool names are consistently in lowercase snake_case, but the pattern varies: imperative verbs for some (authenticate, query) vs. verb_noun for others (get_quickstart, list_databases). Minor inconsistency but still readable.
Four tools is well-scoped for a database query server: authentication, discovery, execution, and reference. Each tool earns its place without redundancy.
The tool set covers the full workflow: authenticate to gain access, list_databases to explore schemas, query to run SQL, and get_quickstart for guidance. No obvious gaps for the intended read-only query domain.
Maintenance
Related MCP Connectors
One MCP server for all web data. Scrape, search, crawl, and AI research in one install.
Hosted MCP server for AI-driven data ops. Create apps, manage schemas, and CRUD structured data.
Hosted MCP server for live public-data APIs and Skills for AI agents.
MCP server for building and testing AI agents with multi-model experimentation and insights.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceMCP server giving AI agents real-time web search, page scraping, company intelligence, email discovery, local lead generation, and a persistent knowledge graph. Pay only for what you use, no subscriptions.7 npmMIT
- FlicenseAqualityDmaintenanceMCP server that acts as a research intelligence agent, converting natural language queries into SQL or full-text search against Hacker News and arXiv data, and returning answers with supporting evidence.4-
- AlicenseNot gradedqualityDmaintenanceMCP server providing a single query tool to execute SQL against GetDot Database, enabling schema discovery, data loading, and querying via natural language.Apache 2.0
- AlicenseNot gradedqualityDmaintenanceAn MCP server providing tools for read-only SQL queries, REST API calls, sandboxed Python execution, and resource lifecycle management, enabling multi-step research and data tasks.MIT