MCP Multi-Agent Server

A multi-domain multi-agent system served over MCP (Model Context Protocol) with FastMCP. A set of specialized domain agents (email, CRM, calendar, customer support, helpdesk, reporting) are exposed as MCP tools and coordinated to automate business workflows, with a bilingual Streamlit dashboard for observability.

Part of the SunnyLab build series. Sanitized public showcase — credentials and infrastructure identifiers removed; configure your own .env.

What it demonstrates

• Multi-agent architecture over MCP / FastMCP — domain agents as composable, policy-routed tools
• Enterprise integrations — Gmail, Salesforce, and other services behind a service layer
• Streamlit dashboard (Korean / English) for runs and observability
• Cloud-native delivery — Docker, docker-compose, Cloud Build, GitHub Actions (project/VM values are placeholders)

Architecture

MCP client (Claude Desktop / Cursor / custom)
        │  MCP
        ▼
FastMCP server ── routes ──► domain agents
   ├─ email      ├─ crm        ├─ calendar
   ├─ cs         ├─ helpdesk   └─ report
        │
        ▼
 service layer (Gmail / Salesforce / …)   →   Streamlit dashboard (KR/EN)

See mcp_server/ for agents, tools, and services.

Tech stack

Python · MCP / FastMCP · Salesforce & Gmail integrations · Streamlit · Docker / docker-compose · Google Cloud Build · GitHub Actions

Project structure

mcp_server/      # agents, tools, services (MCP/FastMCP)
dashboard.py     # Streamlit dashboard (KR)
dashboard_en.py  # Streamlit dashboard (EN)
tests/           # unit tests
cloudbuild.yaml  # Cloud Build (placeholders)
docker-compose.yml · Dockerfile
.env.example     # required env vars (no real keys)

Setup

cp .env.example .env      # fill in your own keys (OPENAI/Google/Salesforce …)
pip install -r requirements.txt
# run the MCP server (see mcp_server/) and the dashboard:
streamlit run dashboard.py

Note

Public portfolio showcase. Credential files, tokens, and infra identifiers (GCP project, VM IP) were removed before publishing; CI/deploy files use placeholders and require your own configuration.

---

SunnyLab — building agentic AI in public · Medium @sunnylabtv · YouTube @sunnylabtv