financy

financy

MCP server that exposes banking data (connections, accounts, balances, transactions) and agent skills, allowing AI agents to query and refresh financial data via stdio.

Category
访问服务器

README

financy

Your Open-Finance / Financy banking data — connections, accounts, balances, transactions — in the terminal, with machine-first output for scripting, an embedded MCP server for agents, and agent skills in the box.

npx financy status

Install

npm install -g financy     # or just use npx financy <command>

Node 20+ is the only prerequisite.

Requirements

The CLI talks to your Open-Finance data through the Financy API, which is a paid feature. To use it you must:

  1. Register for a Financy account at open-finance.ai.
  2. Subscribe to a paid plan (Starter or Pro) — the data API is not available on the free plan (every data command returns exit code 4).
  3. Copy your clientId, clientSecret, and userId from the Financy app → Settings → API.

Setup

financy setup                 # interactive prompts (explains where to get the values)
financy setup --no-input      # read FINANCY_CLIENT_ID / _SECRET / _USER_ID from env (agents/CI)

setup validates the credentials against the live API before saving, so an unregistered/free account is caught immediately (exit 3 for bad credentials, 4 for an ineligible plan).

Commands

financy status                       Are my connections fresh? one-line-per-bank rollup
financy connections list|get <id>    Bank/card connections and their fetch state
financy accounts list|get <id>       Accounts with balances (securities embedded)
financy transactions list|get <id>   Transactions with --from/--to/--account/--type filters
financy categories                   The category taxonomy (English + Hebrew)
financy providers list|branches      Reference data: banks and branches
financy refresh                      Trigger an on-demand refresh of all connections (20 credits)
financy config                       Show resolved endpoints + credential sources (secret masked)
financy skills list|install          Agent skills bundled with the CLI
financy mcp                          Run the embedded MCP server (stdio)

Debugging tip: financy config shows which endpoints and credentials are in effect (and whether each came from env or the config file) without ever printing the secret. Set FINANCY_DEBUG=1 to have the raw API response bodies printed to stderr.

Every command takes --json for a stable machine-readable envelope ({data, count, nextPage} for lists, {data} for single resources; errors as {error:{code,message}} on stderr). List commands take --limit, --cursor, and --all (auto-paginate). Exit codes: 0 ok · 1 unexpected · 2 usage · 3 auth · 4 plan · 5 credits · 6 not-found · 7 api.

Agent skills

Skills ship inside this package, so they can never drift from the CLI version they drive. Install them into a project's .claude/skills/ directory:

financy skills list             # what's in the box
financy skills install --all    # install every skill here
financy skills install freshness-check --dir ~/work/analysis
Skill What it teaches an agent
financy-setup Onboarding a user end to end: install, find the credentials, save them without ever echoing the secret, verify, and explain the paid-plan requirement on exit 4.
freshness-check Reading financy status --json, deciding whether the data is current enough to answer with, and confirming the 20-credit cost before triggering a refresh.

Skills are plain SKILL.md files — read them under skills/ before you install them.

MCP server (for AI agents)

financy mcp runs a stdio Model Context Protocol server exposing the command surface 1:1 as verb_noun tools (list_connections, get_status, refresh_connections, …) with the same {data, …} envelopes. Add it to Claude Code:

claude mcp add financy -- npx financy mcp

Credentials resolve exactly as the CLI does (config file or FINANCY_* env vars); an unconfigured server returns a structured NOT_CONFIGURED error from every tool. refresh_connections costs 20 credits — its tool description tells agents to confirm with the user first.

Updating

financy update

Detects how it was installed and does the right thing: a global install runs npm install -g financy@latest; under npx it reminds you that npx always runs the latest; as a project dependency it defers to your project's package manager.

Releasing

  1. Update CHANGELOG.md (move items out of Unreleased into the new version).
  2. npm version <patch|minor|major> — bumps package.json and creates a vX.Y.Z tag.
  3. git push --follow-tags.
  4. The Release workflow (.github/workflows/release.yml) runs on the tag: it re-runs lint/typecheck/test/build, verifies the tag matches package.json, and publishes to npm with --provenance (via the NPM_TOKEN secret and id-token trusted publishing). The published version shows its provenance attestation on npm.

Before tagging, run the manual staging smoke test (not part of CI) with a real paid-org credential:

npm run build
FINANCY_CLIENT_ID=… FINANCY_CLIENT_SECRET=… FINANCY_USER_ID=… \
FINANCY_AUTH_URL=… FINANCY_API_URL=… FINANCY_CHAT_URL=… FINANCY_AUDIENCE=… \
npm run smoke

Spec & design

License

Apache-2.0

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选