Integrating trip planning into AI conversations with TripGo’s MCP server
In the age of AI assistants, users expect to move seamlessly between information gathering and practical tasks. We’re excited to show how TripGo’s transport intelligence now integrates directly with AI assistants through MCP (Model Context Protocol).
Why This Matters
Imagine asking your AI assistant about attractions in Sydney, then following up with “How do I get from my hotel to the Opera House using public transport?” AI assistants will answer queries like this already with reasonable sounding suggestions, but public transport stops they suggest might not exist, and if they do, the schedules might not be right, and of course they lack real-time information or information about temporary closures or route changes due to events. So to be reliable they need to tap into actual real-world data. This is where external systems like TripGo’s powerful trip planning capabilities can come in to enhance the AI assistants capabilities with accurate, real-time transport information including:
- Multi-modal route planning combing public transport, walking, cycling, taxi, and driving
- Nearby transit stops and departures
- Accessibility information
- Real-time information
The marriage of TripGo’s transport expertise with an AI assistant’s world knowledge creates a truly helpful travel companion that understands both what you want to do and how to get there.
MCP: Specific-purpose tools for general purpose assistants
Model Context Protocol (MCP) is an open standard that serves as a universal connector between AI models and external tools or services. It’s described as “USB-C for AI integrations” – a common language that allows AI assistants to interact with diverse applications through a standardized interface.
Before MCP, integrating AI with external tools required custom code for each combination – a fragmented and inefficient approach. MCP solves this by providing a single protocol for all these interactions, dramatically simplifying the integration process. With MCP, developers can expose their application’s functionality once, making it instantly accessible to any AI that speaks the protocol.
Currently, MCP is natively supported by Anthropic’s Claude (in both the Desktop app and API) and developer-focused AI assistants like Cursor and Windsurf. The ecosystem is growing rapidly, with tools being created for applications ranging from creative software (Blender, Figma, Ableton) to development tools (GitHub, Unity) and productivity services (Zapier’s integration brings access to 8,000+ apps). OpenAI announced they are working on adding MCP support to ChatGPT, and so did Google.
As the standard matures, we can expect more AI platforms and applications to adopt MCP, creating a rich ecosystem where AI assistants can seamlessly access and control a wide variety of tools based on your natural language requests.
How It Works: A Simple Remote MCP Server
We’ve built a straightforward remote MCP server that connects to the TripGo API, exposing key functionality as tools that AI assistants can use. The server provides three primary capabilities:
1. Trip Planning: Find optimal routes between any two points using various transportation modes
2. Location Search: Discover transit stops and mobility options near specific coordinates
3. Departure Information: Get real-time departure schedules from transit stops
The implementation leverages Cloudflare Workers for global deployment and the Model Context Protocol to standardize communication between our server and AI assistants.
Technical Implementation
Under the hood, our MCP server is built using the Cloudflare Workers platform and the Model Context Protocol TypeScript SDK. This serverless approach means our MCP server is quick to iterate on, scales automatically, and requires no infrastructure management.
Cloudflare provides a nice and concise guide for setting up a remote MCP server, which we followed in our approach.
We implemented a small number of tools, which each are defined with parameters, descriptions, and execution logic that calls the TripGo API.
The key aspect is annotating the parameters with descriptions to provide context for the AI assistant and appropriate formats to use. Say, the modes defined in the TripGo API are not obvious and without them the agent might try “transit” or “taxi”, but by specifying the explicit values a frontier model like Claude 3.7 or 4 is able to understand the options; or, if the LLM makes an invalid request, the list of valid modes will be returned and the LLM will generally handle the error.
const tripgoRoutingTool = {
name: "tripgo_routing",
description: "Plan a trip between two locations with various transportation modes", parameters: z.object({
fromLat: z.number().describe("Latitude of the origin location"), // Other location parameters...
departureTime: z
.string()
.optional()
.describe("ISO datetime string for departure time"),
modes: z
.array(z.enum(["pt_pub", "cy_bic", "me_car", "ps_tax", "wa_wal"])) .optional()
.describe("Transportation modes to include."),
// Additional parameters...
}),
execute: async (params) => {
// Call TripGo API and process the results
}
};
With the output, we took care to modify the very detailed responses from the TripGo API to something more compact and easier to “read” for the LLM. That helps the LLM understand the data better, generate more accurate responses to the user’s prompts, and minimises token usage.
Experience It Today with Claude Desktop
You can add TripGo’s trip planning capabilities to Claude Desktop for Windows and macOS. Best of all, this works even without a paid Claude plan! Here’s how to set it up:
1. Open Claude Desktop and go to Settings > Developer
2. Select “Edit Config” and add a TripGo entry like this:
{
"mcpServers": {
"TripGo": {
"command": "npx",
"args": [
"mcp-remote",
"https://tripgo-mcp-server.skedgo-account.workers.dev/sse" ]
}
}
}1. Save the config and restart Claude Desktop
2. You can now ask questions like: “I’m staying near Central Station in Sydney and want to visit the Opera House tomorrow morning. What are my public transport options?”
Claude will ask for your location and destination details, call our TripGo MCP server behind the scenes, and present your trip options in a conversational format. We’ve optimized the experience to be friendly on the low conversation limits of the free plan.
Building AI Agents with TripGo and Claude’s API
Beyond the desktop experience, developers can integrate TripGo’s MCP tools directly into their AI applications using Claude’s API. This enables you to build powerful AI agents or assistants that can provide accurate and real-time trip planning information.
Anthropic recently introduced their MCP connector feature, which allows you to connect to remote MCP servers directly from the Messages API. This means you can create applications that leverage both Claude’s intelligence and TripGo’s transportation expertise without implementing a separate MCP client.
Here’s how to add TripGo to your Claude API calls:
{
"model": "claude-sonnet-4-20250514",
"max_tokens": 1000,
"messages": [
{"role": "user", "content": "I need to get from Central Station to Bondi Beach tomo ],
"mcp_servers": [
{
"type": "url",
"url": "https://tripgo-mcp-server.skedgo-account.workers.dev/sse", "name": "tripgo-mcp"
}
]
}This integration is perfect for creating specialized travel assistants, transportation planning systems, or enhancing existing applications with intelligent mobility features.
Want to integrate TripGo’s capabilities into your own AI applications? Check out our GitHub repository for technical details or contact us to learn more about our API and MCP offerings.
What’s Next for TripGo and AI Integration
This is just the beginning of how transport intelligence can enhance AI conversations. We’re exploring several exciting expansions:
Autonomous Planning Agents: Imagine setting up an assistant that automatically checks your commute each morning and sends notifications about delays
Proactive Notifications: Get alerts about real-time transport conditions impacting your commute
Calendar Integration: Assistants that understand your schedule and proactively suggest optimal travel times, providing a new take on our existing Agenda feature
By bringing transport intelligence directly into AI conversations, we’re making trip planning more natural and accessible.
Get in touch with us if you want to know more: join our Slack team on this page.