taxsort-mcp

taxsort-mcp

Personal tax transaction classification (US Schedule A & C) MCP server that imports bank CSVs, classifies transactions with Claude AI, and provides tools for review, override, summaries, and cost tracking.

Category
访问服务器

README

taxsort-mcp

MCP server for personal tax transaction classification (US Schedule A & C). Built on FastMCP, monetized via Tollbooth DPYC™ Lightning micropayments.

Don't Pester Your Customer™ (DPYC™) — API monetization for Entrepreneurial Bitcoin Advocates

Inspired by The Phantom Tollbooth by Norton Juster, illustrated by Jules Feiffer (1961).

What It Does

Import bank CSV exports (SoFi, Chase, Schwab, US Bank, PayPal, Coinbase), classify transactions into IRS tax categories using Claude AI, review and override classifications, generate tax summaries grouped by IRS line item, and track API usage costs.

Tools

All domain tools take proof: str for DPYC authentication.

Sessions

Tool Description
create_session Create a tax session for a year
list_sessions List your sessions
get_session Session details + stats
session_heartbeat Presence heartbeat; returns who else is active in the session

Verification & Unlock

Tool Description
request_npub_proof Start npub verification via Secure Courier DM
receive_npub_proof Complete npub verification (reply to the DM to prove npub ownership)
verify_passphrase Verify a passphrase to unlock a timed-out session
request_unlock Request a session unlock after timeout
check_unlock Check if the unlock response is valid

Import

Tool Description
import_csv Import CSV — idempotent, preserves user edits
get_import_stats Import sources, date ranges, ambiguous counts

Transactions

Tool Description
get_transactions Filter by category, month, account, date range, search, unclassified-only
get_transactions_paged Server-side filtered, grouped, sorted, paginated transactions
save_classifications Bulk write classifications from the FE (JSON array)
delete_classification Remove a single classification, reverting to unclassified
clear_transactions Delete all transactions and classifications for a session
delete_account_transactions Delete all transactions for a specific imported account
reset_classifications Delete all classifications but keep the imported transactions
get_amount_neighbors Fetch transactions with the same amount within ±N days (duplicate detection)

Accounts

Tool Description
get_accounts List all accounts in a session with types and transaction counts
set_account_type Set an account's type: bank, card, investment, or loan

Summaries

Tool Description
get_summary Grouped spending totals by IRS line, category, month, account

Rules

Tool Description
get_rules Get patron's classification rules
save_rule Create a description-pattern + optional amount-filter → subcategory rule
delete_rule Delete a rule by ID
apply_rules Re-apply all rules to unclassified transactions
count_rule_matches Live preview of how many transactions match a rule pattern

Custom Categories

Tool Description
get_custom_categories Get custom categories defined by this user
save_custom_category Add a custom category/subcategory
delete_custom_category Delete a custom category

Sharing

Tool Description
create_share_token Generate a token to share session with spouse
load_share_token Load a shared session via token

AI Advisors

Tool Description
ask_advisor Ask the Financial Advisor about using TaxSort
ask_tax_researcher Ask the Tax Code Researcher about IRS provisions

API Usage & Feedback

Tool Description
report_api_usage Report Anthropic API usage from FE classification for cost tracking
get_api_usage_stats Get aggregated API usage statistics for cost analysis
get_anthropic_key Get the Anthropic API key for FE-driven classification
get_github_token Get the GitHub token for creating feedback issues
create_feedback_issue Create a GitHub issue for bug reports, feature requests, or feedback
list_feedback_issues List feedback issues submitted by this patron

Standard DPYC™ (from tollbooth-dpyc wheel)

check_balance, purchase_credits, check_payment, check_price, service_status, request_credential_channel, receive_credentials, forget_credentials, oracle_how_to_join, oracle_about, oracle_lookup_member, oracle_network_advisory, oracle_get_tax_rate

Server-Side Pagination

get_transactions_paged performs filtering, grouping, sorting, and pagination entirely on the server. This is more efficient than fetching all transactions and processing client-side.

  • group_by: none, category, subcategory, account, month
  • group_sort: asc or desc (controls group ordering)
  • sort_col + sort_dir: row ordering within each group
  • page + page_size: zero-indexed pagination

Amount Filter Expressions

Rules (save_rule, count_rule_matches) support compound filters with amount_operator and amount_value:

Operator Meaning
lt Less than (<-95 matches debits over $95)
lte Less than or equal
gt Greater than (gte 50 matches amounts $50+)
gte Greater than or equal
eq Exact match (!33 style negation not supported — use neq)
neq Not equal

Examples: <-95 (debits exceeding $95), [0..10) (range via two rules), gte 50.

Dual Classification

  • save_classifications — bulk write from the frontend. Accepts a JSON array of {id, category, subcategory, ...} objects. Used during AI classification and manual review.
  • delete_classification — single revert. Removes one classification, returning the transaction to unclassified state.

Architecture

  • Pure MCP — no custom REST endpoints. Horizon exposes Streamable HTTP (JSON-RPC 2.0)
  • One env varTOLLBOOTH_NOSTR_OPERATOR_NSEC. Neon database provisioned automatically by Authority. BTCPay and Anthropic API key delivered via Secure Courier
  • Per-npub isolation — each patron's transactions, rules, and sessions are keyed by their Nostr public key
  • NeonVault — all persistence via Neon's HTTP SQL API (httpx, no asyncpg)
  • Tollbooth DPYC™ — pre-funded Lightning balances, Authority-certified purchases, constraint-driven pricing

Getting Started

pip install -e ".[dev]"
export TOLLBOOTH_NOSTR_OPERATOR_NSEC=nsec1...
python server.py

Schema migration runs automatically on startup.

Operator Onboarding

  1. Generate a Nostr keypair, set TOLLBOOTH_NOSTR_OPERATOR_NSEC
  2. Deploy to Horizon
  3. Register with Authority: register_operator(npub=..., service_url=...)
  4. Deliver credentials via Secure Courier:
    {
      "btcpay_host": "https://btcpay.example.com",
      "btcpay_api_key": "...",
      "btcpay_store_id": "...",
      "anthropic_api_key": "sk-ant-..."
    }
    

DPYC™ Social Contract

Part of the DPYC™ Social Contract — a network of monetized MCP services where AI agents pay for what they use via Bitcoin Lightning.

Other Operators in the Network

Repo What it does
tollbooth-dpyc Shared Python SDK — vault, auth, pricing, Lightning, Nostr identity
dpyc-community Governance registry — membership, advisories, threat model
dpyc-oracle Community concierge — free onboarding and member lookup
tollbooth-authority Certification backbone — Schnorr-signed certificates
tollbooth-sample Sample Operator (canonical template) — weather stats reference
tollbooth-pricing-studio iOS pricing-model editor / operator console
cypher-mcp Monetized graph answers — named Cypher templates over Neo4j/AuraDB
schwab-mcp Charles Schwab brokerage data — positions, quotes, options, orders
thebrain-mcp TheBrain knowledge graph — thoughts, links, attachments
excalibur-mcp X (Twitter) posting — social media automation with OAuth2
optionality-mcp Options analytics — brokerage-data operator
tollbooth-oauth2-collector OAuth2 callback handler — advocate service
tollbooth-shortlinks URL shortener utility

License

Apache License 2.0. Copyright 2026 Lonnie VanZandt.

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