Skip to main content
Glama

Installation

pip install databento-mcp

Quick Start

  1. Get your API key from Databento

  2. Configure your MCP client (see setup guides below)

  3. Start querying market data through your AI assistant

Setup Guides

Claude Desktop

Add to your configuration file:

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

{
  "mcpServers": {
    "databento": {
      "command": "databento-mcp",
      "env": {
        "DATABENTO_API_KEY": "your-api-key"
      }
    }
  }
}

GitHub Copilot CLI

Add the server to your Copilot CLI configuration:

gh copilot config set mcp-servers '{
  "databento": {
    "command": "databento-mcp",
    "env": {
      "DATABENTO_API_KEY": "your-api-key"
    }
  }
}'

Or add to your ~/.config/gh-copilot/config.yml:

mcp-servers:
  databento:
    command: databento-mcp
    env:
      DATABENTO_API_KEY: your-api-key

See GitHub Copilot CLI MCP documentation for more details.

ChatGPT (via Developer Mode)

ChatGPT supports MCP servers through Developer Mode.

  1. Enable Developer Mode in ChatGPT settings

  2. Add an MCP server with the following configuration:

{
  "name": "databento",
  "command": "databento-mcp",
  "env": {
    "DATABENTO_API_KEY": "your-api-key"
  }
}

See OpenAI Developer Mode documentation for detailed setup instructions.

Features

Historical Data

  • Retrieve trades, OHLCV bars, market depth, and more

  • Support for all Databento schemas (trades, mbp-1, mbp-10, ohlcv-*, etc.)

  • Cost estimation before query execution

  • Smart data summaries with statistics

Live Data

  • Real-time market data streaming

  • Configurable stream duration

  • Multiple schema support

File Operations

  • Read/write DBN format files

  • Export to Apache Parquet

  • Convert between formats

Batch Processing

  • Submit large-scale batch jobs

  • Monitor job status

  • Download completed files

Reference Data

  • Symbol metadata and definitions

  • Symbology resolution

  • Dataset discovery

  • Publisher information

Quality & Performance

  • Smart caching with configurable TTL

  • Data quality analysis

  • Connection pooling

  • Comprehensive metrics

Available Tools

Tool

Description

health_check

Check API connectivity and server status

get_historical_data

Retrieve historical market data

get_live_data

Stream real-time market data

get_cost

Estimate query cost before execution

get_symbol_metadata

Get instrument definitions and mappings

search_instruments

Search for symbols with wildcards

list_datasets

List available Databento datasets

list_schemas

List available data schemas

resolve_symbols

Convert between symbology types

submit_batch_job

Submit batch data download

list_batch_jobs

List batch job status

get_batch_job_files

Get batch job download info

cancel_batch_job

Cancel pending batch job

download_batch_files

Download completed batch files

read_dbn_file

Parse and read DBN files

get_dbn_metadata

Get DBN file metadata

write_dbn_file

Write data to DBN format

convert_dbn_to_parquet

Convert DBN to Parquet

export_to_parquet

Query and export to Parquet

read_parquet_file

Read Parquet files

get_session_info

Get trading session info

list_publishers

List data publishers

list_fields

List schema fields

get_dataset_range

Get dataset date range

list_unit_prices

Get pricing information

analyze_data_quality

Analyze data quality issues

quick_analysis

Comprehensive symbol analysis

get_account_status

Server status and metrics

get_metrics

Performance metrics

clear_cache

Clear API response cache

Configuration

Environment Variable

Description

Default

DATABENTO_API_KEY

Databento API key (required)

-

DATABENTO_DATA_DIR

Restrict file operations to directory

Current directory

DATABENTO_LOG_LEVEL

Logging level (DEBUG, INFO, WARNING, ERROR)

INFO

DATABENTO_METRICS_ENABLED

Enable metrics collection

true

Common Datasets

Dataset

Description

GLBX.MDP3

CME Globex (ES, NQ, CL futures)

XNAS.ITCH

Nasdaq TotalView

XNYS.PILLAR

NYSE

DBEQ.BASIC

Consolidated US equities

OPRA.PILLAR

US options

IFEU.IMPACT

ICE Futures Europe

Common Schemas

Schema

Description

trades

Individual trades

ohlcv-1m

1-minute OHLCV bars

ohlcv-1h

1-hour OHLCV bars

ohlcv-1d

Daily OHLCV bars

mbp-1

Top of book

mbp-10

10-level order book

tbbo

Top bid/offer

definition

Instrument definitions

Development

# Clone repository
git clone https://github.com/deepentropy/databento-mcp.git
cd databento-mcp

# Install with dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Format code
black src/
ruff check src/

License

MIT License

Available Tools

30 tools
analyze_data_qualityA

Analyze data quality and detect issues in market data.

Detects:

  • Time gaps in data

  • Price outliers (>3 standard deviations)

  • Null values and missing data

  • Duplicate records

Returns:

  • Quality score (0-100)

  • List of issues and warnings

  • Detailed breakdown of problems

Example:

  • First retrieve data with get_historical_data

  • Then analyze_data_quality(dataset="GLBX.MDP3", symbols="ES.FUT", start="2024-01-15", end="2024-01-15")

Inputs: dataset (required), symbols (required), start (required), end (required), +2 more.

cancel_batch_jobC

Cancel a pending or processing batch job.

Parameters:

  • job_id (required) - The batch job ID to cancel

Returns:

  • Success/failure status

  • Final job state

  • Message explaining the result

Example: cancel_batch_job(job_id="JOB-12345")

Inputs: job_id (required).

clear_cacheC

Clear the API response cache

Inputs: expired_only.

convert_dbn_to_parquetC

Convert a DBN file to Parquet format

Inputs: input_path (required), output_path, compression.

download_batch_filesA

Download completed batch job files to a local directory.

Parameters:

  • job_id (required) - The batch job ID

  • output_dir (required) - Directory to save downloaded files

  • overwrite (optional, default: false) - Whether to overwrite existing files

Returns:

  • List of downloaded files with paths and sizes

  • Total download size

  • Success/failure status for each file

Example: download_batch_files(job_id="JOB-12345", output_dir="/data/downloads")

Inputs: job_id (required), output_dir (required), overwrite.

export_to_parquetC

Query historical data and export directly to Parquet format

Inputs: dataset (required), symbols (required), schema (required), start (required), +3 more.

get_account_statusB

Get comprehensive server status and account information.

Returns:

  • Server uptime and health

  • Cache statistics (hit rate, entries)

  • API usage metrics

  • Tool usage breakdown

  • Memory and performance stats

Example: get_account_status() for a full server overview

Inputs: no named parameters.

get_batch_job_filesC

Get download information for a completed batch job

Inputs: job_id (required).

get_costA

Estimate the cost of a historical data query before executing it

Inputs: dataset (required), symbols (required), schema (required), start (required), +1 more.

get_dataset_rangeB

Get the available date range for a dataset

Inputs: dataset (required).

get_dbn_metadataB

Get only the metadata from a DBN file without reading all records

Inputs: file_path (required).

get_historical_dataA

Retrieve historical market data for symbols from Databento.

Examples:

  • Get ES futures trades: dataset="GLBX.MDP3", symbols="ES.FUT", start="2024-01-15", end="2024-01-15", schema="trades"

  • Get AAPL OHLCV bars: dataset="XNAS.ITCH", symbols="AAPL", start="2024-01-01", end="2024-01-31", schema="ohlcv-1m"

  • Get BTC order book: dataset="GLBX.MDP3", symbols="BTC.FUT", schema="mbp-10"

Tips:

  • Use explain=true to preview query without executing

  • Use force_refresh=true to bypass cache

  • Start with small date ranges and use limit parameter

Inputs: dataset (required), symbols (required), start (required), end (required), +4 more.

get_live_dataC

Subscribe to real-time market data for a limited duration

Inputs: dataset (required), symbols (required), schema, duration.

get_metricsB

Get server performance metrics and usage statistics.

Returns:

  • Server uptime

  • API call counts

  • Cache hit/miss rates

  • Per-tool latency and success rates

Example: get_metrics(reset=false) to view current stats

Inputs: reset.

get_session_infoC

Identify the current trading session based on time

Inputs: timestamp.

get_symbol_metadataC

Get metadata for symbols including symbology mappings and instrument definitions

Inputs: dataset (required), symbols (required), start (required), end.

health_checkA

Check the health and connectivity of the Databento API. Use this to diagnose connection issues or verify the server is working properly.

Inputs: verbose.

list_batch_jobsB

List all batch jobs with their current status

Inputs: states, since, limit.

list_datasetsB

List all available datasets from Databento

Inputs: no named parameters.

list_fieldsC

List fields available for a specific schema

Inputs: schema (required), encoding.

list_publishersC

List data publishers with their details

Inputs: dataset.

list_schemasB

List all available data schemas from Databento.

Returns:

  • List of all available schemas with descriptions

  • Includes trades, OHLCV bars, order book, and reference data schemas

Example: list_schemas()

Inputs: no named parameters.

list_unit_pricesB

Get current pricing information per dataset/schema combination.

Parameters:

  • dataset (optional) - Filter by dataset name (e.g., "GLBX.MDP3")

Returns:

  • List of unit prices showing cost per GB or per record

  • Helps understand pricing before querying

Example: list_unit_prices(dataset="GLBX.MDP3")

Inputs: dataset.

quick_analysisA

One-call comprehensive analysis of a symbol.

Combines: metadata + cost estimate + sample data + trading session info + data quality check.

Example:

  • quick_analysis(dataset="GLBX.MDP3", symbol="ES.FUT", date="2024-01-15")

  • quick_analysis(dataset="XNAS.ITCH", symbol="AAPL", date="2024-01-15", schema="ohlcv-1m")

Returns:

  • Symbol metadata and instrument info

  • Cost estimate for full-day data

  • Sample of recent trades/bars

  • Current trading session context

  • Data quality assessment

Inputs: dataset (required), symbol (required), date (required), schema.

read_dbn_fileC

Read and parse a DBN file, returning the records as structured data

Inputs: file_path (required), limit, offset.

read_parquet_fileC

Read a Parquet file and return the data

Inputs: file_path (required), limit, columns.

resolve_symbolsC

Resolve symbols between different symbology types

Inputs: dataset (required), symbols (required), stype_in (required), stype_out (required), +2 more.

search_instrumentsC

Search for instruments in a dataset

Inputs: dataset (required), start (required), symbols.

submit_batch_jobC

Submit a batch data download job for large historical datasets

Inputs: dataset (required), symbols (required), schema (required), start (required), +4 more.

write_dbn_fileC

Write historical data query results directly to a DBN file

Inputs: dataset (required), symbols (required), schema (required), start (required), +3 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3.1/5.0

    Scored across 30 tools

    Disambiguation3/5

    Most tools have distinct purposes, but there is some overlap that could cause confusion. For example, get_account_status and get_metrics both provide server performance information, and analyze_data_quality and quick_analysis both assess data quality. However, descriptions help differentiate them, and the majority of tools target unique operations.

    Naming Consistency4/5

    Tool names follow a consistent verb_noun pattern throughout, such as get_historical_data, list_datasets, and submit_batch_job. There are minor deviations like clear_cache (verb_adjective) and health_check (noun_noun), but the overall naming is predictable and readable.

    Tool Count2/5

    With 30 tools, the count is too high for the server's purpose of market data access and analysis. Many tools feel redundant or overly granular, such as separate tools for reading DBN and Parquet files or multiple batch job operations, which could overwhelm agents and indicate poor scoping.

    Completeness4/5

    The tool surface is largely complete for market data operations, covering data retrieval, analysis, batch processing, and metadata. Minor gaps exist, such as no direct tool for updating or deleting data, but agents can work around this given the domain's focus on read-heavy operations.

    Related MCP Connectors