OneTick MCP Server

Enterprise-grade MCP server for OneTick tick data analytics. Provides 18 tools across 4 categories, 6 workflow commands, and 6 domain skills — covering equities, futures, FX, options, and indices from 200+ global exchanges.

Quick Start

1. Install

git clone https://github.com/your-org/onetick-mcp.git
cd onetick-mcp
pip install -e .

2. Set Credentials

export ONETICK_CLIENT_ID=your_client_id
export ONETICK_CLIENT_SECRET=your_client_secret

Or copy .env.example to .env and fill in your credentials.

3. Connect to Claude

Choose your platform below.

---

Platform Setup

Claude Code (CLI)

Register the MCP server so Claude Code can use all 18 tools:

# Add to your current project
claude mcp add onetick -- onetick-mcp

# Or with explicit credentials
claude mcp add onetick \
  --env ONETICK_CLIENT_ID=your_client_id \
  --env ONETICK_CLIENT_SECRET=your_client_secret \
  -- onetick-mcp

# Verify it's registered
claude mcp list

This creates a .mcp.json in your project root. To register across all projects instead:

claude mcp add --scope user onetick -- onetick-mcp

Direct mode (registers all 18 tools upfront instead of 3 meta-tools — uses more tokens but skips the discovery step):

claude mcp add onetick -- onetick-mcp --direct

Claude Desktop

Add to your Claude Desktop configuration file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "onetick": {
      "command": "uv",
      "args": [
        "run",
        "--directory",
        "/path/to/onetick_mcp",
        "onetick-mcp"
      ],
      "env": {
        "ONETICK_CLIENT_ID": "your_client_id",
        "ONETICK_CLIENT_SECRET": "your_client_secret"
      }
    }
  }
}

Replace /path/to/onetick_mcp with the actual path to this repository.

---

Using Skills and Commands

The server ships with 6 skills (domain knowledge) and 6 commands (step-by-step workflows) that tell Claude how to chain the MCP tools into complete analyses. These work together: commands define what to do, skills provide how to interpret the results.

Skills (Domain Knowledge)

Skills are loaded from the skills/ directory. Each skill teaches Claude a domain — what metrics matter, how to interpret them, and what output format to produce.

Skill	Domain	When to Use
tca-analysis	Execution benchmarking	Trading costs, slippage, VWAP shortfall, cost decomposition
market-microstructure	Liquidity & price formation	Order book depth, bid-ask dynamics, buy/sell pressure
intraday-analytics	Volume & momentum	Volume profile, VWAP deviation, unusual activity detection
volatility-analysis	Risk measurement	Realized vol, vol regimes, historical percentile comparison
execution-quality	Fill assessment	Interval-level benchmark comparison, best/worst fill windows
market-overview	Daily briefing	Price action, volume, volatility, spreads across symbols

Commands (Workflow Orchestration)

Commands are loaded from the commands/ directory. Each command chains 3-5 tool calls into a complete analysis with a defined workflow.

Command	What It Does	Tool Chain
/analyze-tca	Transaction Cost Analysis	CALC_VWAP -> CALC_SPREAD_STATS -> CALC_TRADE_STATS -> GET_BARS
/analyze-microstructure	Market microstructure	GET_BOOK_SNAPSHOT -> CALC_BOOK_IMBALANCE -> CALC_SPREAD_STATS
/analyze-intraday	Intraday activity profile	CALC_VWAP -> CALC_TRADE_STATS -> GET_BARS
/analyze-volatility	Realized volatility analysis	CALC_VOLATILITY -> GET_DAILY_BARS -> CALC_SPREAD_STATS -> CALC_TRADE_STATS
/analyze-execution	Execution quality assessment	CALC_VWAP -> CALC_TRADE_STATS -> CALC_SPREAD_STATS -> GET_BARS
/market-overview	Market snapshot	GET_DAILY_BARS -> CALC_VWAP -> CALC_TRADE_STATS -> CALC_VOLATILITY -> CALC_SPREAD_STATS

How to Use in Claude Code

Once the MCP server is registered, you can use commands and ask questions naturally:

# Run a workflow command
/analyze-tca AAPL 2024-01-15 09:30:00 2024-01-15 16:00:00

# Ask natural language questions (Claude picks the right tools)
"What's the VWAP for MSFT today?"
"Show me the order book for TSLA"
"Morning briefing for AAPL, MSFT, GOOGL"
"How volatile is AMZN compared to the last 20 days?"

# Run multi-symbol analysis
/market-overview AAPL,MSFT,GOOGL,AMZN

How Skills and Commands Work Together

Each command references its corresponding skill. For example, /analyze-tca uses the tca-analysis skill for domain expertise:

1. Command defines the workflow: which tools to call, in what order, with what parameters
2. Skill provides interpretation: what the numbers mean, how to classify results, what output format to use
3. MCP Tools execute the computation: deterministic analytics on OneTick's C++ engine

This separation means you can also ask free-form questions. Claude will use the skill knowledge to pick the right tools and interpret results, even without invoking a command explicitly.

---

Packaging as a Plugin

To distribute the server + skills + commands as a self-contained Claude Code plugin:

1. Create Plugin Manifest

Create .claude-plugin/plugin.json:

{
  "name": "onetick",
  "description": "OneTick market data analytics — tick data, TCA, microstructure, volatility analysis across 200+ global exchanges",
  "version": "0.2.0",
  "author": {
    "name": "OneTick"
  }
}

2. Create Plugin MCP Config

Create .mcp.json at the repo root:

{
  "mcpServers": {
    "onetick": {
      "type": "stdio",
      "command": "onetick-mcp",
      "env": {
        "ONETICK_CLIENT_ID": "${ONETICK_CLIENT_ID}",
        "ONETICK_CLIENT_SECRET": "${ONETICK_CLIENT_SECRET}"
      }
    }
  }
}

3. Test the Plugin

claude --plugin-dir /path/to/onetick_mcp

When packaged as a plugin, commands are namespaced:

/onetick:analyze-tca AAPL 2024-01-15 09:30:00 2024-01-15 16:00:00

Plugin Directory Structure

onetick_mcp/
├── .claude-plugin/
│   └── plugin.json            <- Plugin manifest
├── .mcp.json                  <- MCP server config (auto-starts with plugin)
├── commands/                  <- Workflow commands (become slash commands)
│   ├── analyze-tca.md
│   ├── analyze-microstructure.md
│   ├── analyze-intraday.md
│   ├── analyze-volatility.md
│   ├── analyze-execution.md
│   └── market-overview.md
├── skills/                    <- Domain knowledge (loaded automatically)
│   ├── tca-analysis/SKILL.md
│   ├── market-microstructure/SKILL.md
│   ├── intraday-analytics/SKILL.md
│   ├── volatility-analysis/SKILL.md
│   ├── execution-quality/SKILL.md
│   └── market-overview/SKILL.md
├── src/                       <- MCP server implementation
│   ├── server.py
│   ├── config.py
│   ├── response.py
│   └── tools/
│       ├── registry.py
│       ├── meta_tools.py
│       ├── data_retrieval.py
│       ├── metadata.py
│       ├── analytics.py
│       └── sql.py
├── tests/
├── CONNECTORS.md              <- Complete tool reference
├── pyproject.toml
└── .env.example

---

MCP Tools Reference

Progressive Discovery (Default)

The server exposes only 3 meta-tools by default, reducing token usage by ~88%:

Meta-Tool	Purpose
TOOL_LIST	List all 18 tools with brief descriptions (~1,000 tokens)
TOOL_GET	Full schema for specific tool(s) (~200 tokens per tool)
TOOL_CALL	Execute a tool by name with JSON arguments

Workflow: TOOL_LIST -> identify relevant tools -> TOOL_GET for schemas -> TOOL_CALL with arguments.

Use --direct mode to register all 18 tools upfront (no meta-tools, but ~8,000 tokens upfront).

Market Data Retrieval (8 tools)

Tool	Description	Key Parameters
GET_TICK_DATA	Raw tick data (trades, quotes, NBBO) for a single symbol	symbol, tick_type, database, start, end, max_rows
GET_BARS	OHLC/VWAP/TWAP bars at configurable intervals	symbol, bar_type, interval, database
GET_DAILY_BARS	End-of-day OHLCV with corporate action adjustment	symbol, start_date, end_date, adjusted
GET_MULTI_SYMBOL	Data for 2+ symbols in parallel	symbols, data_type, bar_type, interval
GET_BOOK_SNAPSHOT	Point-in-time order book reconstruction	symbol, timestamp, max_levels
GET_BOOK_TIMESERIES	Order book snapshots at regular intervals	symbol, start, end, interval
GET_CORPORATE_ACTIONS	Splits, dividends, mergers, adjustment factors	symbol, start_date, end_date
GET_STATIC_DATA	Reference data (name, currency, ISIN) or auction prices	symbol, data_type

Metadata & Discovery (4 tools)

Tool	Description
LIST_DATABASES	All available databases by region and asset class
GET_DATABASE_INFO	Tick types, date range, schema for a specific database
SEARCH_SYMBOLS	Find symbols by pattern (SQL LIKE: 'AAPL', 'AA%', '%GOLD%')
LIST_VENUES	All supported exchanges by region and asset class

Analytics & Computation (5 tools)

All computations are deterministic, executed on OneTick's C++ engine.

Tool	Description	Formula
CALC_VWAP	Single aggregate VWAP for a time range	SUM(Price*Volume) / SUM(Volume)
CALC_SPREAD_STATS	Bid-ask spread statistics per interval	Spread = ASK - BID
CALC_BOOK_IMBALANCE	Order book buy/sell pressure	(BidVol - AskVol) / (BidVol + AskVol)
CALC_VOLATILITY	Realized volatility from trade data	StdDev(log returns), annualized
CALC_TRADE_STATS	Trade flow: count, volume, VWAP, avg size per interval	Aggregated from trade ticks

SQL (1 tool)

Tool	Description
EXECUTE_SQL	Run OneTick SQL SELECT queries. Table format: DATABASE.TICK_TYPE

See CONNECTORS.md for complete parameter specifications and optimization guidance.

---

Usage Examples

Quick Lookups

"What is the current price of AAPL?"
-> GET_TICK_DATA (symbol='AAPL', tick_type='TRD', max_rows=1)

"EUR/USD rate right now"
-> GET_TICK_DATA (symbol='EUR/USD', database='GLOBAL_FX', tick_type='QTE', max_rows=1)

"What's the DJIA at?"
-> GET_TICK_DATA (database='DJ_INDICES', max_rows=1)

Daily / Historical Data

"AAPL daily chart for this month"
-> GET_DAILY_BARS (symbol='AAPL', start_date='2026-04-01', end_date='2026-04-30')

"MSFT historical prices adjusted for splits"
-> GET_DAILY_BARS (symbol='MSFT', adjusted=True)

Intraday Bars

"5-minute OHLC bars for AAPL today"
-> GET_BARS (symbol='AAPL', bar_type='ohlc', interval='5min')

"Compare 5-min bars for AAPL, MSFT, GOOGL"
-> GET_MULTI_SYMBOL (symbols='AAPL,MSFT,GOOGL', data_type='bars', interval='5min')

Analytics

"What's the VWAP for AAPL today?"
-> CALC_VWAP (symbol='AAPL', start='2026-04-08 09:30:00', end='2026-04-08 16:00:00')

"Is there buying pressure in TSLA?"
-> CALC_BOOK_IMBALANCE (symbol='TSLA')

"Realized volatility for GOOGL"
-> CALC_VOLATILITY (symbol='GOOGL', interval='5min')

Workflow Commands

"Run a TCA for CSCO from 9:30 to 12:00 on Jan 3, 2024"
-> /analyze-tca chains: CALC_VWAP -> CALC_SPREAD_STATS -> CALC_TRADE_STATS -> GET_BARS

"Analyze AAPL's market microstructure"
-> /analyze-microstructure chains: GET_BOOK_SNAPSHOT -> CALC_BOOK_IMBALANCE -> CALC_SPREAD_STATS

"Morning briefing for AAPL, MSFT, GOOGL"
-> /market-overview chains: GET_DAILY_BARS -> CALC_VWAP -> CALC_TRADE_STATS -> CALC_VOLATILITY -> CALC_SPREAD_STATS

SQL Queries

SELECT SYMBOL_NAME, SUM(SIZE) AS VOLUME
FROM US_COMP.TRD
WHERE SYMBOL_NAME = 'AAPL'
  AND TIMESTAMP >= '2024-01-15 09:30:00 America/New_York'
GROUP BY SYMBOL_NAME

---

Tool Selection Guide

Question Type	Use This Tool	NOT This
"What is the price of X?"	GET_TICK_DATA (max_rows=1)	GET_DAILY_BARS
"X daily chart"	GET_DAILY_BARS	GET_BARS or GET_TICK_DATA
"VWAP for X" (single number)	CALC_VWAP	GET_BARS
"VWAP bars every 5min" (time series)	GET_BARS (bar_type=vwap)	CALC_VWAP
"Volume today"	CALC_TRADE_STATS	GET_TICK_DATA
"Bid-ask spread"	CALC_SPREAD_STATS	GET_TICK_DATA
"Show order book"	GET_BOOK_SNAPSHOT	CALC_BOOK_IMBALANCE
"Buying/selling pressure"	CALC_BOOK_IMBALANCE	GET_BOOK_SNAPSHOT
"What databases exist?"	LIST_DATABASES	LIST_VENUES
"What exchanges exist?"	LIST_VENUES	LIST_DATABASES
"Compare multiple symbols"	GET_MULTI_SYMBOL	Multiple GET_TICK_DATA calls

---

Supported Databases

Database	Asset Class	Region	Examples
US_COMP	Equities	US	AAPL, MSFT, GOOGL, TSLA
CME	Futures	US	ES (S&P), CL (crude oil), GC (gold), NG (nat gas)
GLOBAL_FX	FX	Global	EUR/USD, GBP/JPY, USD/JPY
LSE	Equities	EU	VOD, BP, HSBA
XETRA	Equities	EU	SIE, SAP, ALV
EURONEXT	Equities	EU	AI, MC, SAN
EUREX	Futures	EU	FESX, FGBL
SP_INDICES	Indices	US	SPX, RUT
DJ_INDICES	Indices	US	INDU (DJIA)
CBOE_IDX	Indices	US	VIX
US_OPTIONS	Options	US	OPRA consolidated
CA_COMP / TSX	Equities	CA	RY, TD, BNS

OneTick cloud demo databases use _SAMPLE suffix (e.g., US_COMP_SAMPLE). The server resolves this automatically — if US_COMP is not found, it tries US_COMP_SAMPLE.

---

Requirements

• Python 3.10+
• OneTick Cloud API credentials (ONETICK_CLIENT_ID and ONETICK_CLIENT_SECRET)
• onetick-py[webapi] package (installed automatically)
• Valid OneTick data entitlements for the databases you want to access

Credential Setup

Obtaining Credentials

1. Log in to your OneTick Cloud account at cloud.onetick.com
2. Navigate to API settings or contact your OneTick administrator
3. Generate a client ID and client secret for API access

Configuration Methods

Method	Best For	How
Environment variables	Development	export ONETICK_CLIENT_ID=...
.env file	Local use	Copy .env.example to .env
Claude Desktop config	Claude Desktop	Add to claude_desktop_config.json
CLI arguments	Quick testing	onetick-mcp --client-id X --client-secret Y

OneTick Connection Details

• REST endpoint: https://rest.cloud.onetick.com:443
• Auth endpoint: https://cloud-auth.parent.onetick.com/realms/OMD/protocol/openid-connect/token
• Authentication: OAuth2 client_credentials flow (automatic)

Running Tests

# Tool selection validation (7 tests)
python tests/test_tool_selection.py

# Workflow/skill structure validation (9 tests)
python tests/test_workflow_validation.py

# Independent query optimization validation (10 rules, 100 queries)
python tests/test_independent_queries.py

License

MIT