🔗 MCP API

Introduction

Use the CoinAPI Flat Files MCP server to discover buckets, browse dataset prefixes, and enumerate downloadable flat files through self-describing tools.

Flat Files MCP

The CoinAPI Flat Files API is available through a hosted Model Context Protocol (MCP) server. It exposes a compact set of self-describing tools for browsing buckets and file paths from MCP clients such as Cursor, Claude Desktop, or custom agent runtimes.

Flat Files are served from two buckets: coinapi (hourly partitions and legacy daily data) and coinapi-daily-tail (previous day's daily data only, 24-hour retention). List both with storage_list_buckets before browsing prefixes.

Server endpoint

https://mcp.flatfiles.coinapi.io/mcp

Use your regular CoinAPI API key in the X-CoinAPI-Key header.

Only the /mcp endpoint is supported for this service.

What is available through MCP?

The Flat Files MCP server currently exposes 2 tools:

AreaWhat you can do
Bucket discoveryList the storage buckets available to your API key and inspect owner metadata.
Prefix and object discoveryBrowse dataset prefixes, date partitions, and concrete .csv.gz objects using S3-style listings.

Typical discovery flow

Most integrations follow this order:

  1. Call storage_list_buckets to list the available buckets: coinapi and coinapi-daily-tail.
  2. Call storage_list_objects with bucket: "coinapi" to enumerate top-level dataset prefixes such as T-TRADES/, T-QUOTES/, T-LIMITBOOK_FULL/, or T-OHLCV_ACTIVE_CONSOLIDATED/.
  3. For the previous day's daily files, call storage_list_objects with bucket: "coinapi-daily-tail" and a dataset prefix such as T-TRADES/.
  4. Call storage_list_objects again with a deeper prefix such as T-TRADES/ in coinapi to discover hourly partitions like T-TRADES/D-2026040712/.
  5. Continue narrowing the prefix until you reach concrete object keys ending in .csv.gz.
  6. Use the discovered object key with the standard S3 API when you want to download the file contents.

Live-tested notes

  • The current MCP listing returns two accessible buckets: coinapi and coinapi-daily-tail.
  • Root-level listing for coinapi currently returns T-TRADES/, T-QUOTES/, T-OHLCV_ACTIVE_CONSOLIDATED/, and T-LIMITBOOK_FULL/.
  • Root-level listing for coinapi-daily-tail returns Trades, Quotes, and Full Limit Order Book prefixes, but not OHLCV. This bucket is a rolling 24-hour window for the previous day's daily files only and is not a historical archive.
  • storage_list_objects returns folder-like values in commonPrefixes, while contents stays empty until you browse a concrete prefix level.
  • A live listing for T-TRADES/ in coinapi currently returns hourly partitions such as T-TRADES/D-2026040712/, T-TRADES/D-2026040711/, and T-TRADES/D-2026040710/.
  • A listing for coinapi-daily-tail exposes a single daily partition for yesterday's date, for example T-TRADES/D-20260609/.
  • MCP is best suited for discovery of valid bucket names and object paths before using your regular download workflow.

What MCP is best for

This MCP server is best suited for:

  • Interactive discovery of bucket structure and dataset families
  • Agent workflows that need to inspect available prefixes before building download paths
  • Enumerating object metadata such as file key, last modification time, and size

If you need to download the file bytes, continue with the standard S3 API after discovering the object key through MCP.

Start here

Service StatusGitHub SDK