AgenticBooks MCP Server
Financial data infrastructure for AI agents. Connect to a startup's books to read live P&L and bank balances, review and reclassify transactions, manage the chart of accounts, and connect banking sources.
README
AgenticBooks MCP Server
Financial data infrastructure for AI agents. Connect Claude (or any MCP client) directly to a startup's books — read live P&L and bank balances, review and reclassify transactions, manage the chart of accounts, and connect banking sources — over a secure, OAuth-authenticated remote MCP server.
This is a hosted (remote) MCP server. You don't install or self-host it — you connect your MCP client to the AgenticBooks endpoint and authenticate with your AgenticBooks account. This repository is the public listing and connection guide; the server itself runs as managed infrastructure.
- Website: https://agenticbooks.com
- MCP endpoint:
https://mcp.agenticbooks.ai/mcp - Transport: Streamable HTTP
- Auth: OAuth 2.0 (Claude "Custom Connector" flow via Clerk) or an
ab_API key
What it does
AgenticBooks ingests webhooks from Stripe, Mercury, RevenueCat and other sources, normalises and classifies the events into double-entry accounting, and syncs to QuickBooks/Xero. This MCP server exposes that ledger to an AI agent so the agent can answer financial questions and take bookkeeping actions on the user's behalf — every write is audit-trailed and scoped to the caller's organisation.
Typical prompts once connected:
- "What's our P&L for last month?"
- "How much cash do we have across all bank accounts right now?"
- "Show me transactions that still need review, grouped by vendor."
- "Reclassify the Vercel charges as Hosting, and remember that for next time."
Connect
Claude (Desktop / Web) — Custom Connector (OAuth)
- Open Settings → Connectors → Add custom connector.
- Enter the MCP server URL:
https://mcp.agenticbooks.ai/mcp - Complete the OAuth sign-in when prompted. You'll authenticate with your AgenticBooks account; access is scoped to your organisation.
See Anthropic's guide to remote custom connectors: https://support.claude.com/en/articles/11175166-get-started-with-custom-connectors-using-remote-mcp
API key (any MCP client over HTTP)
If your client doesn't support the OAuth connector flow, authenticate with an
ab_ API key (create one in the AgenticBooks dashboard) as a bearer token:
{
"mcpServers": {
"agenticbooks": {
"transport": "http",
"url": "https://mcp.agenticbooks.ai/mcp",
"headers": {
"Authorization": "Bearer ab_your_api_key_here"
}
}
}
}
The two auth modes are independent — OAuth and ab_ keys both work; you only
need one.
Tools
All tools operate within the authenticated organisation. Reads are non-destructive; writes are recorded in an append-only audit trail.
Reporting & balances
| Tool | What it does |
|---|---|
get_financial_summary |
One-call snapshot: current-month P&L, balances across all connected providers, unreviewed-event count, and ledger sync health. |
get_pnl_report |
Profit & loss report over a date range, built from classified events. |
get_account_balance |
Current balances across the org's connected money providers. |
get_bank_balances |
Bank holdings, one line per connected bank (Mercury, etc.). |
get_audit_log |
Reads the append-only audit trail, newest first. |
Review & classification
| Tool | What it does |
|---|---|
get_unreviewed_events |
Lists classified events the rules engine couldn't confidently categorise. |
get_pending_by_counterparty |
Summarises the review queue clustered by counterparty and direction. |
approve_classification |
Approves one event from the review queue by assigning its account. |
reclassify_entry |
Changes the account on an already-posted ledger entry (audit-trailed). |
Counterparty rules (the learned vendor map)
| Tool | What it does |
|---|---|
create_counterparty_rule |
Creates/updates a persistent rule mapping a counterparty to an account. |
list_counterparty_rules |
Lists the org's learned counterparty → account rules. |
update_counterparty_rule |
Re-points an existing rule at a different account (affects future events). |
delete_counterparty_rule |
Removes a rule; future events from that counterparty stop auto-classifying. |
Chart of accounts
| Tool | What it does |
|---|---|
list_chart_accounts |
Lists the org's chart of accounts (code, name, type, active). |
add_chart_account |
Adds a new account; the system auto-assigns the code. |
rename_chart_account |
Renames an account (the code never changes). |
set_chart_account_active |
Deactivates or reactivates a chart account. |
Periods & integrations
| Tool | What it does |
|---|---|
close_period |
Runs the month-end close, including unrealized FX revaluation. |
reimport_historical |
Queues a historical import over a chosen window. |
connect_mercury |
Connects the org to Mercury via a read-only, scoped API token. |
disconnect_mercury |
Disables the org's Mercury integration (audit-trailed). |
Security
- Every tool call is authenticated (OAuth access token or
ab_API key) and scoped to a single organisation. - All writes are recorded in an append-only audit trail.
- Banking credentials are stored per-organisation and encrypted at rest; they are never accepted as, or exposed via, MCP tool arguments.
Links
- Website — https://agenticbooks.com
- Support — https://agenticbooks.com (contact via dashboard)
AgenticBooks is a hosted service. This repository documents how to connect; it is not the server source.
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。