fisc
An open-source MCP server that enables AI agents to read, populate, and manage tax returns in Canadian professional tax software like Taxprep and DT Max through a vendor-neutral concept layer.
README
fisc
Open-source MCP server for professional tax software.
Connect AI agents to Taxprep, DT Max, and other Canadian tax preparation packages through the Model Context Protocol.
What is this?
Canadian accounting firms prepare tax returns in software like Taxprep (Wolters Kluwer) and DT Max (Thomson Reuters). Those systems were built around professional tax workflows, not around one common interface for AI agents. fisc is an open interoperability layer that aims to give agents a consistent way to read, populate, and manage returns across tax software.
Your AI agent
↓ MCP
fisc
↓ adapters
Taxprep · DT Max · ...
Why open-source?
We do not think the connector should be the lock-in.
Accounting firms should be able to connect AI to the tax software they already trust through an interface that is inspectable, portable, and open to contributions. A shared interoperability layer also makes it easier for developers and software vendors to add adapters without rebuilding the same plumbing in private.
Coalesc is building its product above this layer: the engagement workflow, document intelligence, evidence, orchestration, review controls, and the reasoning that decides what should happen next. Open the rails; compete on the intelligence and workflow.
Status
Early development. fisc is not yet functional. We're building the Taxprep adapter first via the CCH iFirm Taxprep Web API, then researching the best supported path for DT Max.
Planned tools
| Tool | Description |
|---|---|
create_return |
Create a new tax return for a taxpayer |
set_field |
Set a value in a return (e.g. employment income, medical expenses) |
get_field |
Read a value from a return |
list_forms |
List available forms in a return |
get_diagnostics |
Retrieve validation diagnostics |
list_returns |
List existing returns |
Planned adapters
- Taxprep (CCH iFirm Taxprep Web API) — in progress
- DT Max (Thomson Reuters) — integration path under research
- More to come
Architecture
fisc uses a vendor-neutral tax concept layer. Instead of asking an agent to understand software-specific cell IDs, it can work with semantic concepts:
// What your agent says:
await client.callTool("set_field", {
return_id: "2026-john-smith",
concept: "employment_income",
value: 82400,
source: "t4_acme.pdf"
});
// An adapter translates that concept to the tax software's native representation.
Concept-to-vendor mappings are maintained per tax year, software, and return type.
Getting started
Note: fisc is not yet functional. Star the repo to follow progress.
# Clone
git clone https://github.com/coalesc/fisc.git
cd fisc
# Install
npm install
# Configure (when adapters are ready)
cp .env.example .env
# Add your CCH iFirm API credentials
# Run
npm start
Project structure
src/
index.ts # MCP server entry point
concepts/ # Vendor-neutral tax concept definitions
adapters/
taxprep/ # CCH iFirm Taxprep adapter
dtmax/ # DT Max adapter (planned)
tools/ # MCP tool implementations
Contributing
We welcome contributions — especially:
- New adapters for professional tax software
- Concept mappings for additional return types (T2, T3, T5013)
- Tax year updates to mappings
- Safer tool and review patterns for AI-assisted tax workflows
See CONTRIBUTING.md for guidelines.
Who builds this?
fisc is built by Coalesc, which is building an end-to-end workspace where accountants and AI prepare, review, and move client engagements forward together. We plan to use fisc as an interoperability layer between that workspace and the tax software firms already use.
License
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。