clickhouse-mcp
Read-only MCP server for large-scale cross-sectional quant analysis of US stock market data (1.49 billion rows) using ClickHouse.
README
clickhouse-mcp
A read-only Model Context Protocol (MCP) server that queries a USA-market ClickHouse warehouse — 1.49 billion rows of 1-minute bars across 1,388 symbols (1991–2026) plus L1 aggregates and L2 materialised technical indicators — and exposes it to MCP-aware agents for large-scale, cross-sectional quant analysis that single-symbol REST MCP servers cannot do.
It is the 6th read-only server in the kevinkda MCP ecosystem (schwab-marketdata / schwab-positions / sec-edgar / polygon-news / yfinance), and deliberately stays independent because its authentication (ClickHouse host/user/password) and data shape (columnar deep history) differ from the REST-based siblings.
Tools (7, all read-only)
| Tool | Purpose |
|---|---|
get_ohlcv |
OHLCV bars for one symbol over a date range at 1m/5m/15m/1h/1d/1w |
get_indicators |
One materialised / runtime indicator series (ma20, rsi14, macd_hist, …) |
screen_stocks |
Full-market technical-indicator scan (e.g. "today's RSI<30 oversold list") |
get_correlation_matrix |
Pairwise return-correlation matrix for 2–50 symbols (Pearson / Spearman) |
run_safe_sql |
Disabled by default raw-SQL escape hatch (single read-only SELECT only) |
health_check |
Local probe + optional ClickHouse reachability check |
get_server_info |
Version + tool list (offline) |
Quickstart
git clone https://github.com/kevinkda/clickhouse-mcp
cd clickhouse-mcp
uv sync --extra dev
cp .env.example .env # then fill in your read-only ClickHouse connection
Configure the connection in .env (see Security):
CLICKHOUSE_MCP_HOST=your-clickhouse-host
CLICKHOUSE_MCP_HTTP_PORT=8123
CLICKHOUSE_MCP_USER=mcp_readonly
CLICKHOUSE_MCP_PASSWORD=...
CLICKHOUSE_MCP_DATABASE=usa
Run it:
uv run clickhouse-mcp
Register with an MCP client
{
"mcpServers": {
"clickhouse": {
"command": "uv",
"args": ["run", "clickhouse-mcp"],
"cwd": "/absolute/path/to/clickhouse-mcp",
"env": {
"CLICKHOUSE_MCP_HOST": "your-clickhouse-host",
"CLICKHOUSE_MCP_HTTP_PORT": "8123",
"CLICKHOUSE_MCP_USER": "mcp_readonly",
"CLICKHOUSE_MCP_PASSWORD": "...",
"CLICKHOUSE_MCP_DATABASE": "usa"
}
}
}
}
Security
This server touches financial data, so it is hardened accordingly. See
docs/SECURITY.md and
docs/THREAT_MODEL.md.
- Dedicated read-only ClickHouse account — point
CLICKHOUSE_MCP_USERat an account created withreadonly = 1andGRANT SELECTonly, never an admin user. Defence in depth: every query is also issued with the ClickHouse session settingreadonly=1plusmax_execution_time/max_result_rowsguardrails. - Parameterised, never concatenated —
get_ohlcv/get_indicators/screen_stocks/get_correlation_matrixvalidate every symbol / indicator / date through anchored regexes + allow-lists and bind them as ClickHouse query parameters. User input never lands in the SQL string. run_safe_sqldisabled by default — gated behindCLICKHOUSE_MCP_ALLOW_RAW_SQL=true; even when enabled it enforces SELECT-only single statements, rejects DDL/DML + comments, forces aLIMIT, and inheritsreadonly=1+ the resource guardrails.- SSRF-safe — the ClickHouse host is read from env at startup and is never derived from tool input; no argument can redirect the outbound connection.
- Credentials redacted — every exception runs its text through
redact_secrets, so a password / DSN can never leak viarepr(exc)or logs.
Quant use cases unlocked
CH is the "full-market + deep-history + batch" engine; the REST MCP siblings are "real-time + fundamentals + news". Together they close the loop:
- Full-market technical scan — RSI/MACD/Bollinger/MA signals across all 1,388 symbols (single-symbol MCP cannot scan the market).
- Cross-sectional factor ranking — value/momentum/quality scoring across the universe on a columnar store.
- Multi-symbol correlation matrix — portfolio / pairs prerequisite; CH batch read beats fetching symbols one at a time.
- Cross-source quant playbook — CH historical signal → schwab real-time price → sec-edgar fundamentals → polygon news.
Development
bash scripts/local-ci.sh # ruff + mypy + bandit + pip-audit + pytest (100% cov)
Tests never touch a live ClickHouse — clickhouse_connect is mocked.
License
MIT — see LICENSE.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.