Build a Google Calendar AI Agent in 60 seconds with Arcade.dev's MCP Servers
Whatβs in the Box
When you deploy the OAP stack, you get:
- LangGraph: agent orchestration layer (graph-based state handling)
- Arcade.dev: secure tool execution and OAuth abstraction via MCP
- Prebuilt agents: starter templates with tools attached
- MCP compatibility: lets agents access MCP tools across services
Agent Creation Workflow
Creating a new agent involves:
- Selecting a base template
- Customize the agent
- Select the model
- Set the system prompt
- Set model parameters like the temperature and token limit
- Attaching tools
- Choose from built-in tools like Gmail, Google Calendar, Slack, etc.
- Arcade.dev tools are discoverable and scoped by default
- Tools use Model Context Protocol (MCP) and are executed remotely via Arcade.dev, with real user auth.
- Attach RAG
- Create specialized agents with access to specific documents and files that it can use as reference
- Saving + initializing agent runtime
Example: Calendar Agent
Config:
- Model: gpt-4o-mini
- Set the temperature parameter to a relatively low value to get more consistent responses: 0.3
- Tools:
- Google Create Event
- Google List Events
- Google List Calendars
Agent behavior:
- Accepts freeform input via chat
- Maps user intent to tool calls
- Authenticates via just-in-time OAuth
- Calls Google Calendar APIs securely
- Returns structured results into LangGraph pipeline
Auth Flow (via Arcade)
Arcade handles end-user authentication via secure OAuth 2.0 flows with:
- Scoped access (least privilege)
- Zero token exposure to LLMs
- High-quality API integrations for services like Gmail, Slack, Google Docs, and Salesforce
Tokens are exchanged behind the scenes (never in view of the LLM). Auth happens only when a tool is needed β not before.
Try It Now
- π Explore the LangChain Demo
- π GitHub: langchain-ai/open-agent-platform
- π Start with Arcade Tools β
Frequently Asked Questions
What is the LangChain Open Agent Platform? Itβs a platform for building, configuring, and deploying AI agents that can call real tools using secure, scoped authentication β powered by LangGraph and Arcade.dev.
Can I use Gmail or Google Calendar with LangChain agents? Yes! With Arcade.dev, agents can safely securely access Gmail, Calendar, Slack, and other APIs on your behalf.
What is MCP and why does it matter? MCP (Model Context Protocol) is an open spec that defines how agents call tools. The OAP stack uses it to make tools composable, streamable, and agent-safe.
Do I need to host my own MCP server? No. Arcade.dev provides hosted support for MCP, but you can also run your own. Itβs configurable per deployment.