FDIC BankFind MCP Server

An MCP (Model Context Protocol) server that provides access to the FDIC BankFind Suite API, giving LLMs programmatic access to public data on FDIC-insured financial institutions.

Features

• No API key required — The FDIC BankFind API is publicly available
• 11 tools covering BankFind datasets plus a built-in comparative analysis tool
• Flexible ElasticSearch-style filtering on all endpoints
• Pagination support for large result sets
• Dual transport: stdio (local) or HTTP (remote)
• Structured tool output via structuredContent for MCP clients
• Built-in analytical batching for multi-bank comparisons
• Short-lived response caching for repeated analytical prompts

Available Tools

| Tool | Description | |------|-------------| | fdic_search_institutions | Search FDIC-insured banks and savings institutions | | fdic_get_institution | Get details for a specific institution by CERT number | | fdic_search_failures | Search failed bank records | | fdic_get_institution_failure | Get failure details for a specific institution | | fdic_search_locations | Search branch/office locations | | fdic_search_history | Search structural change events (mergers, name changes, etc.) | | fdic_search_financials | Search quarterly Call Report financial data | | fdic_search_summary | Search annual financial summary data | | fdic_search_sod | Search Summary of Deposits (branch-level deposit data) | | fdic_search_demographics | Search quarterly demographics and market-structure data | | fdic_compare_bank_snapshots | Compare two reporting snapshots across banks and rank growth/profitability changes |

Two tools are convenience lookups rather than separate BankFind datasets:

• fdic_get_institution wraps the institutions dataset
• fdic_get_institution_failure wraps the failures dataset

One tool is a server-side analysis helper:

• fdic_compare_bank_snapshots batches roster lookup, financial snapshots, and optional demographics snapshots inside the MCP server so complex trend prompts do not require many separate tool calls
• It supports both snapshot comparisons and timeseries analysis across a quarterly range
• It computes derived efficiency and balance-sheet metrics and assigns insight tags for notable growth patterns

Filter Syntax

All tools support ElasticSearch query string syntax for filtering:

# Exact match
NAME:"Chase Bank"

# State filter
STNAME:"California"

# Numeric range (assets in $thousands)
ASSET:[1000000 TO *]

# Date range
FAILDATE:[2008-01-01 TO 2010-12-31]

# Combine with AND / OR
STNAME:"Texas" AND ACTIVE:1 AND ASSET:[500000 TO *]

# Exclude
!(BKCLASS:NM OR BKCLASS:N)

Field names vary by dataset. For example:

• institutions commonly uses fields like STNAME, ASSET, ACTIVE
• failures commonly uses fields like FAILDATE, COST, RESTYPE
• locations commonly uses fields like STALP, CITY, BRNUM
• sod commonly uses fields like STALPBR, CITYBR, DEPSUMBR

Use the tool description or fields parameter to tailor queries to the dataset you are calling.

Installation

npm install
npm run build

Development

npm run typecheck
npm test
npm run build

CI And Releases

• Pull requests should target main
• Continuous integration runs on pushes to main and on pull requests targeting main
• Package publish is reserved for semantic version tags like v1.2.3
• Release tags must point at a commit already on main

To prepare a release:

npm version patch
git push origin main --follow-tags

Publishing from GitHub Actions is intended to use npm trusted publishing via GitHub Actions OIDC rather than a long-lived NPM_TOKEN.

Usage

CLI bundle

The build produces two outputs:

• dist/index.js: CLI entrypoint for stdio/HTTP server execution
• dist/server.js: import-safe library bundle exporting createServer, createApp, and main

stdio (for Claude Desktop / local MCP clients)

node dist/index.js

Claude Desktop config (~/Library/Application\ Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "fdic": {
      "command": "node",
      "args": ["/path/to/fdic-mcp-server/dist/index.js"]
    }
  }
}

HTTP server

TRANSPORT=http PORT=3000 node dist/index.js

The MCP endpoint will be available at http://localhost:3000/mcp.

For direct HTTP MCP clients, include:

• Content-Type: application/json
• Accept: application/json, text/event-stream

The streamable HTTP transport expects both response types to be accepted.

Example Queries

Find all active banks in North Carolina with assets over $1 billion:

filters: STNAME:"North Carolina" AND ACTIVE:1 AND ASSET:[1000000 TO *]

Get the 10 costliest bank failures since 2000:

filters: FAILDATE:[2000-01-01 TO *]
sort_by: COST
sort_order: DESC
limit: 10

Get all branches of a specific bank:

cert: 3511
(fdic_search_locations)

Get quarterly financials for Bank of America for 2023:

cert: 3511
filters: REPDTE:[20230101 TO 20231231]
(fdic_search_financials)

Get demographics history for a bank:

cert: 3511
sort_by: REPDTE
sort_order: DESC
(fdic_search_demographics)

Rank banks by growth across two dates without orchestrating many separate queries:

state: North Carolina
start_repdte: 20211231
end_repdte: 20250630
sort_by: asset_growth
sort_order: DESC
(fdic_compare_bank_snapshots)

Analyze quarterly trends with streaks and derived metrics:

state: North Carolina
start_repdte: 20211231
end_repdte: 20250630
analysis_mode: timeseries
sort_by: asset_growth_pct
sort_order: DESC
(fdic_compare_bank_snapshots)

Complex Prompts

The built-in fdic_compare_bank_snapshots tool is meant to make analysis-heavy prompts performant by batching FDIC calls inside the MCP server. The server also caches repeated state/date lookups for a short period, which helps iterative analysis. Good examples:

• Identify North Carolina banks with the strongest asset growth from 2021 to 2025, then compare whether that growth came with higher deposits, more branches, or better profitability
• Rank a set of banks by deposit growth percentage between two report dates and check which ones also improved ROA or ROE
• Compare current active banks in a state between two snapshots to see which institutions grew while reducing office counts
• Analyze whether rapid asset growth was accompanied by stronger profitability or just balance-sheet expansion
• Analyze quarterly trends from 2021 through 2025 and call out inflection points, sustained asset-growth streaks, or multi-quarter ROA declines
• Identify banks whose deposits-per-office and assets-per-office improved even while total branch counts fell
• Separate banks with branch-supported growth from banks that mainly expanded their balance sheets

The tool can take either:

• state plus optional institution_filters to build a roster
• certs to compare a specific list of institutions directly

Notable outputs from fdic_compare_bank_snapshots:

• point-to-point changes for assets, deposits, net income, ROA, ROE, and office counts
• derived metrics such as assets_per_office_change, deposits_per_office_change, and deposits_to_assets_change
• insight tags such as growth_with_better_profitability, growth_with_branch_expansion, balance_sheet_growth_without_profitability, and growth_with_branch_consolidation
• in timeseries mode, quarterly series plus streak metrics like sustained asset growth and multi-quarter ROA decline

Response Shape

All tools return:

• a human-readable text representation in content[0].text
• machine-readable data in structuredContent

Typical search response shape:

{
  "total": 1,
  "offset": 0,
  "count": 1,
  "has_more": false,
  "institutions": [
    {
      "CERT": 3511,
      "NAME": "Wells Fargo Bank, National Association"
    }
  ]
}

Typical lookup miss response shape:

{
  "found": false,
  "cert": 999999999,
  "message": "No institution found with CERT number 999999999."
}

Data Notes

• Monetary values are in $thousands unless otherwise noted
• The CERT field is the unique FDIC Certificate Number for each institution
• The ACTIVE field: 1 = currently active, 0 = inactive/closed
• Report dates (REPDTE) are in YYYYMMDD format
• The financials endpoint covers 1,100+ Call Report variables — use fields to narrow results
• fdic_search_financials defaults to sort_order: DESC for most-recent-first results
• Most other search tools default to sort_order: ASC
• The demographics endpoint is useful for office counts, metro/micro flags, territory assignments, and related geographic reference attributes
• The demographics dataset is quarterly historical data, not just a current snapshot