agnifolio-mcp
Official MCP server for Agni Folio, a free multi-currency wealth & portfolio tracker. 23 tools for holdings, performance, FIRE planning, crypto P&L, and confirm-gated writes — hosted endpoint with OAuth 2.1, plus a Docker stdio bridge in this repo.
README
Agni Folio MCP Server
Official Model Context Protocol server for Agni Folio — a free, multi-currency wealth & portfolio tracker (stocks, crypto, real estate, fixed deposits, mutual funds, insurance, and loans across 20+ markets).
This is a hosted remote server. There is nothing to install or run — you connect your MCP client to the production endpoint and authenticate with your own Agni Folio account.
- Endpoint:
https://agnifolio.com/mcp(streamable HTTP) - Auth: OAuth 2.1 with Dynamic Client Registration + PKCE — any MCP client can onboard without pre-registration
- Scopes:
agnifolio:read,agnifolio:write(per-user, revocable from Settings → Connected AI) - Server card:
/.well-known/mcp/server-card.json - Agent skills:
/.well-known/agent-skills/index.json - Docs: Connect Your AI · auth.md · llms.txt
Quick connect
Claude (web/desktop): Settings → Connectors → Add custom connector → https://agnifolio.com/mcp → complete the OAuth sign-in.
Cursor / Cline / Continue / other stdio-first clients — bridge via mcp-remote:
{
"mcpServers": {
"agnifolio": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://agnifolio.com/mcp"]
}
}
}
The first call opens a browser window for the OAuth consent flow; tokens are cached locally by mcp-remote.
Public-tools server (open source, in this repo)
This repo also contains a small runnable MCP server for Agni Folio's public tools — no account needed:
| Tool | What it does |
|---|---|
calculate_fire_number |
FIRE number, progress %, years-to-FIRE, Coast FIRE — pure local math |
get_high_interest_products |
Current savings/FD/treasury rates across SG, IN, US (public API) |
find_networth_percentile |
Net-worth percentile by country/age/gender (anonymized public dataset) |
get_agnifolio_info |
About Agni Folio + how to connect to the personal-data server |
npm install && npm run build
node dist/index.js # stdio MCP server
# or via Docker
docker build -t agnifolio-public-mcp .
docker run -i --rm agnifolio-public-mcp
Claude Desktop / Cursor config:
{
"mcpServers": {
"agnifolio-public": {
"command": "node",
"args": ["/path/to/agnifolio-mcp/dist/index.js"]
}
}
}
The public-tools server never sees credentials; personal portfolio data is only available through the hosted OAuth server above.
Tools (23)
Read — portfolio: query_holdings, query_holding_by_symbol, query_account_summary, query_performance_summary, query_classification_warnings
Read — FIRE planning: query_user_fire_status, query_user_coast_fire, query_fire_what_if
Read — crypto & trading: query_crypto_pnl, query_exchange_spot_balances, query_open_positions, query_margin_detail, query_recent_trades
Read — protection & legacy: query_insurance_summary, query_nominees
Write (confirm-gated): add_portfolio_entry, add_transaction, add_insurance_policy, delete_portfolio_entry, parse_text_into_entries, set_crypto_cost_basis, recover_crypto_cost_basis, update_user_fire_goals
Every mutating tool is confirm-before-execute: the change is staged and must be approved inside the Agni Folio app before it takes effect. Every external call — read or write — is recorded in a per-user audit log.
Security model
- Per-user OAuth tokens; the server never sees your client's credentials
- Scoped access (
readvswrite) chosen at consent time - Revocation: one click in Settings → Connected AI kills the grant
- Confirm-gated writes + full audit trail
- Agni Folio is not a brokerage and executes no trades; it is a tracking/analytics tool
About this repository
The server implementation is part of the Agni Folio production backend and is not open source; this repository is the canonical public home for connection docs and issue reports for the MCP surface. Found a problem with a tool? Open an issue or use contact.
推荐服务器
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。
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。