LangGraph Accounting Agent Example
📂 View Source Code This example demonstrates how to use LangGraph's ReAct agent with MCP tools for accounting queries.
Features​
- Framework: LangGraph with
create_react_agent
- Tools: MCP tools converted to LangChain tools via
langchain-mcp-adapters
- Agent Type: ReAct (Reasoning and Acting) pattern
- Model: ChatOpenAI (GPT-4.1-nano)
Setup​
- Install dependencies:
uv sync
- Set up your environment variables (create a
.env
file):
GOOGLE_API_KEY=your_google_api_key_here
GEMINI_API_KEY=${GOOGLE_API_KEY}
Usage​
uv run main.py
Configuration​
The proxy_config.yaml
file defines a mocked Odoo MCP target that provides accounting tools like:
get_expenses
- Retrieve expense recordsget_revenues
- Retrieve revenue recordsget_account_balance
- Get account balances
The configuration uses LLM-generated responses for realistic accounting data simulation.