Overview
The Eyeshot MCP Server allows AI-powered coding assistants to query Eyeshot's API documentation, code samples, and Knowledge Base articles directly from your IDE or chat client.
With the Eyeshot MCP Server, your AI assistant can:
- Answer questions about Eyeshot types, methods, and properties
- Retrieve C# and VB.NET code samples from the official Eyeshot sample library
- Look up how-to articles and troubleshooting guides from the Eyeshot Knowledge Base
- Generate code suggestions grounded in up-to-date Eyeshot documentation
See the Eyeshot MCP Server in action:
Prerequisites
- An active Eyeshot subscription
- An MCP-compatible AI client (see the list below)
Supported AI clients and tools:
- Claude Desktop
- Claude Code (CLI)
- VS Code with GitHub Copilot
- Cursor
- Windsurf
- JetBrains Rider AI
- ChatGPT (Plus, Team, or Enterprise)
- Gemini CLI
- any other MCP-compatible AI assistant
Getting your MCP Token
Each request to the Eyeshot MCP Server must be authenticated with a personal token. To retrieve your token:
- Log in and navigate to your profile page
- Locate the MCP Token row at the bottom of the User panel
- Click the copy icon next to the token value
You can regenerate the token at any time by clicking the refresh icon. After regeneration, update your client configuration with the new value.
Configuration
Claude Desktop
Open Settings > Connectors, click Add custom connector, and enter the following values:
- Name: Eyeshot
-
URL:
https://devdept.com/mcp?key=YOUR_MCP_TOKEN
Replace YOUR_MCP_TOKEN with the token copied from your profile.
Claude Code (CLI)
Run the following command in your terminal:
claude mcp add eyeshot --transport http https://devdept.com/mcp \
--header "Authorization: Bearer YOUR_MCP_TOKEN" \
--scope userReplace YOUR_MCP_TOKEN with the token copied from your profile. Use --scope project instead of --scope user to restrict the server to the current project only.
VS Code (GitHub Copilot)
Create or open .vscode/mcp.json in your workspace root and add the following configuration:
{
"servers": {
"eyeshot": {
"type": "http",
"url": "https://devdept.com/mcp",
"headers": {
"Authorization": "Bearer YOUR_MCP_TOKEN"
}
}
}
}Replace YOUR_MCP_TOKEN with the token copied from your profile. Reload the VS Code window after saving the file.
Cursor / Windsurf / JetBrains Rider AI
Open your client's MCP configuration file and add the following entry under mcpServers:
{
"mcpServers": {
"eyeshot": {
"url": "https://devdept.com/mcp",
"headers": {
"Authorization": "Bearer YOUR_MCP_TOKEN"
}
}
}
}Replace YOUR_MCP_TOKEN with the token copied from your profile.
-
Cursor:
~/.cursor/mcp.json -
Windsurf:
~/.codeium/windsurf/mcp_config.json - JetBrains Rider AI: see Settings > Tools > AI Assistant > Model Context Protocol
Available tools
The Eyeshot MCP Server exposes eight tools across four domains. Your AI assistant selects and chains them automatically based on your question. Every tool accepts an optional version parameter (e.g. "2026"); when omitted, the latest ready version is used.
Getting started
-
eyeshot_get_started: returns a short workflow guide that tells the assistant what Eyeshot is, which tools to use, and in what order. The assistant is expected to call this tool first at the start of any Eyeshot-related task; no manual trigger is needed.
API Documentation
-
eyeshot_docs_search: searches the Eyeshot API documentation by keyword and returns a list of matching types -
eyeshot_docs_get_content: returns the overview and full member list of a specific type -
eyeshot_docs_get_member: returns the complete documentation for a specific member (parameters, return type, example)
Code Samples
-
eyeshot_samples_search: searches the official C# and VB.NET code sample library, with optional platform and language filters -
eyeshot_samples_get_content: returns the full source code of a specific sample
Knowledge Base
-
eyeshot_kb_search: searches the Eyeshot Knowledge Base articles -
eyeshot_kb_get_content: returns the full content of a specific Knowledge Base article
Asking questions that get good answers
The most effective way to use the AI Assistant is to be specific about what you are building and which platform you are targeting. Vague questions get generic answers; concrete questions get working code.
You do not need to know the API in advance. Just describe what you want to achieve.
| Instead of this... | Try this |
|---|---|
| How do I use Eyeshot? | Show me the standard XAML to add an Eyeshot control to display a 3D scene. |
| Load a file. | In Eyeshot, let the user pick a STEP file and load it, then zoom to fit the scene. |
| Change the display mode. | Toggle the active viewport between Rendered and Wireframe when a button is clicked, and update the button label accordingly. |
| Get the selected objects. | In the Eyeshot control, handle the selection change event to show the type and color of each selected entity in a label. |
| Show multiple views. | Split the viewport to show two views side by side: Isometric on the left and Front on the right. |
| Export the model. | Let the user save all entities to a GLB file asynchronously. |
If the first result is not quite right, describe what is wrong specifically. "It doesn't work" leaves the AI guessing. AI models can sometimes take a roundabout approach when answering questions. If that happens, a nudge like "check the Eyeshot MCP tools" is usually enough to get things back on track.
💡 For best results, start from a project that already has the Eyeshot NuGet package installed. This lets the AI focus on the actual task rather than project setup, and keeps token usage low.
 Important
Always review AI-generated code before using it in production. Output quality depends on how the question is phrased, the AI model in use, and the clarity of your existing codebase.
Comments
Does the MCP provide documentation for older versions of Eyeshot or only the latest version?
Hi Magnus,
The MCP server currently supports Eyeshot 2025 and 2026. Older versions aren't included at the moment.
I’ve been meaning to suggest you set something like this up for a while now. But as I didn’t want to bother you, I left it at that. 😅
It’s definitely a really useful tool – thank you very much!!
Hi Andreas,
Next time, don't hesitate: suggestions like this are exactly how we figure out what to build next! Really glad it's useful. 😊
Please sign in to leave a comment.