mcpField
A
Autodesk Product Help MCP

Autodesk Product Help MCP

Remote
help.autodesk.comOther0Updated 1d ago

About Autodesk Product Help MCP

Autodesk Product Help MCP is a read-only server for MCP-compatible AI clients seeking official help and support content across Autodesk products. It supports natural-language searches and product or release filtering without requiring credentials.

Key features

  1. 01Search official help and support content
  2. 02List supported products and releases
  3. 03Filter searches by product and release
  4. 04Return ranked results with source URLs
  5. 05Search in 10 languages and 40 locales

Use cases

  1. 01Find instructions for an Autodesk product task
  2. 02Troubleshoot product errors using official articles
  3. 03Browse documentation for a specific release

MCP support

Exposes two read-only tools to MCP clients: a product catalog lookup and a documentation search. Clients connect to its remote Streamable HTTP endpoint without credentials.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://api.aps.autodesk.com/knowledge/public/v1/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Tools

Details

From the project README

Autodesk Product Help MCP

Give any MCP-compatible AI client read-only access to Autodesk's official Help and Support documentation.

MCP compatible Streamable HTTP No authentication Read-only

Getting started · Tools · Examples · Troubleshooting · Docs


Autodesk® Product Help is a Model Context Protocol (MCP) Server that connects AI agents directly to Autodesk’s official product documentation across 110+ products including Autodesk Fusion®, AutoCAD®, Revit®, Inventor®, Maya®, Civil 3D®, and more. As a secure, read-only interface, it allows AI tools to search, navigate, and retrieve trusted Autodesk help content. This enables support for natural language queries, accelerates onboarding, and provides a reliable foundation for intelligent agents built on authoritative Autodesk product documentation.

Key Features:

  • Ask support questions using natural language.
  • Learn and onboard to Autodesk products faster.
  • Create a foundation for AI agents and automation.
  • Available in 10 languages and 40 locales.

Requirements:

  • No special requirements.

About Autodesk

The world's designers, engineers, builders, and creators trust Autodesk to help them design and make anything. From the buildings we live and work in, to the cars we drive and the bridges we drive over. From the products we use and rely on, to the movies and games that inspire us. Autodesk's Design and Make Platform unlocks the power of data to accelerate insights and automate processes, empowering our customers with the technology to create the world around us and deliver better outcomes for their business and the planet. For more information, visit autodesk.com.

Important

Autodesk hosts and operates this service, so there is nothing to install or run locally. This folder of the ai-integrations repository contains the MCP Registry metadata (server.json) and setup documentation.

Example prompts

Once connected, try asking your assistant:

  • How-to: "In AutoCAD, how do I create and manage layers using Layer Properties Manager?" or "In AutoCAD or Civil 3D, how do I create and assign a custom property to piping components using the Spec Editor?"
  • Troubleshooting: "In Revit, why are my linked CAD files not visible in a view, and how can I troubleshoot the issue?" or "What does AutoCAD error 1603 mean?"
  • Getting started: "I'm new to Revit. Show me a beginner tutorial for placing walls." or "Which Autodesk products does this help cover?"
  • Account and billing: "How do I sign in to my Autodesk account?" or "How do I upgrade my trial to a paid subscription?"

Getting started

Endpoint: https://api.aps.autodesk.com/knowledge/public/v1/mcp

Any MCP client that supports remote servers over Streamable HTTP can connect to this URL. No credentials are needed. A health check is available at https://api.aps.autodesk.com/knowledge/public/health.

Use it in Claude or ChatGPT

Autodesk Product Help is listed in both apps, so you can connect it without any configuration:

Add the server to your client
Claude Code
claude mcp add --transport http autodesk-product-help \
  https://api.aps.autodesk.com/knowledge/public/v1/mcp
Cursor

Add to Cursor

Or add it manually to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project):

{
  "mcpServers": {
    "autodesk-product-help": {
      "url": "https://api.aps.autodesk.com/knowledge/public/v1/mcp"
    }
  }
}
VS Code

Install in VS Code

Or add it manually to .vscode/mcp.json in your workspace:

{
  "servers": {
    "autodesk-product-help": {
      "type": "http",
      "url": "https://api.aps.autodesk.com/knowledge/public/v1/mcp"
    }
  }
}

You can also run MCP: Add Server from the Command Palette, choose HTTP, and paste the endpoint URL.

Windsurf

Add the server to ~/.codeium/windsurf/mcp_config.json, then restart Windsurf:

{
  "mcpServers": {
    "autodesk-product-help": {
      "serverUrl": "https://api.aps.autodesk.com/knowledge/public/v1/mcp"
    }
  }
}
Other clients and SDKs

For other clients, add a remote HTTP server that points to the endpoint URL. The configuration key names vary, so check your client's MCP documentation.

To call the server from your own code, use an official MCP SDK.

Python (pip install "mcp>=2"):

import asyncio

from mcp import ClientSession
from mcp.client.streamable_http import streamable_http_client

MCP_URL = "https://api.aps.autodesk.com/knowledge/public/v1/mcp"

async def main() -> None:
    async with streamable_http_client(MCP_URL) as (read, write):
        async with ClientSession(read, write) as session:
            await session.initialize()
            result = await session.call_tool(
                "search_help_content",
                {"query": "AutoCAD error 1603", "locale": "en_US"},
            )
            structured = result.structured_content or {}
            for doc in structured.get("results", []):
                print(doc["title"], doc["url"])


if __name__ == "__main__":
    asyncio.run(main())

TypeScript / Node.js (npm install @modelcontextprotocol/sdk):

import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StreamableHTTPClientTransport } from "@modelcontextprotocol/sdk/client/streamableHttp.js";

const MCP_URL = "https://api.aps.autodesk.com/knowledge/public/v1/mcp";

const client = new Client({ name: "example", version: "1.0.0" });
await client.connect(new StreamableHTTPClientTransport(new URL(MCP_URL)));

const { structuredContent } = await client.callTool({
  name: "search_help_content",
  arguments: { query: "AutoCAD error 1603", locale: "en_US" },
});

for (const doc of structuredContent?.results ?? []) {
  console.log(doc.title, doc.url);
}

await client.close();
Let your agent set it up

Paste this into your AI coding agent:

Add the Autodesk Product Help MCP server to this environment.
Use the setup documentation at https://github.com/Autodesk/ai-integrations/tree/main/product-help/remote-mcp-server
and the endpoint https://api.aps.autodesk.com/knowledge/public/v1/mcp
(Streamable HTTP, no authentication). Verify the connection with initialize and tools/list.
Verify the connection
curl -sS -X POST https://api.aps.autodesk.com/knowledge/public/v1/mcp \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{
        "protocolVersion":"2025-06-18","capabilities":{},
        "clientInfo":{"name":"curl","version":"1.0"}}}'

A successful response reports serverInfo.name as autodesk-knowledge-public-v1-mcp-server.

Tools

The server exposes two tools. Both are annotated as read-only, non-destructive, and idempotent.

ToolPurpose
get_available_productsList the supported Autodesk products, with their product codes and release codes
search_help_contentSearch Help and Support documentation and return ranked results with source URLs
get_available_products

Returns the product catalog as a list of { product_name, product_code, release_codes } objects. It takes no arguments.

Call it when you need a valid product_code or release_code to narrow a search. Release codes are identifiers such as 2025, not display names such as "AutoCAD 2025". The catalog can change over time, so treat this tool as the authoritative list rather than hardcoding values.

search_help_content

Searches Autodesk documentation and support articles and returns the most relevant documents.

ParameterRequiredDescription
queryYesKeyword-rich search text, 2 to 1500 characters. Include the product, the feature or task, and any exact error codes. Write it in the same language as locale.
localeYesLanguage and region for search and results, for example en_US, de_DE, ja_JP, or zh_CN. The tool accepts regional locale values, several of which can map to the same language (for example de_DE, de_AT, and de_CH are all German). The tool schema lists every accepted value.
product_codeNoLimits results to one product. Must be a value from get_available_products. Preferred over product_name.
product_nameNoAlternative to product_code. If both are set, they must refer to the same product.
release_codeNoLimits results to one release, for example 2025. Requires product_code or product_name, and must be a valid release for that product.
sourcesNosfdc_articles (support knowledge base articles), product_documentation (official product documentation), or both. Searches both if omitted.
max_resultsNoNumber of results to return, 1 to 50. Defaults to 5.

Each result includes title, url, locale, body (full text or an excerpt), media_type, and source. The response also includes total_results.

Example call:

{
  "name": "search_help_content",
  "arguments": {
    "query": "AutoCAD Error 1603 install failed",
    "locale": "en_US",
    "product_code": "ACD",
    "max_results": 3
  }
}
Best practices for agents
  • Call search_help_content directly when the product is already known. Call get_available_products only when you need to look up or validate a product or release code.
  • If you cannot map a product or release to an exact value from get_available_products, leave that filter out rather than guessing.
  • Make query keyword-rich: include the product and the feature or task, keep the user's wording where possible, and include exact error text or codes.
  • Write query in the same language as locale, and take locale from the user's context rather than guessing it.
  • Cite the url of each result you use, so users can read the original article.

Troubleshooting

MCP clients and SDKs handle the HTTP details for you. The HTTP errors below mostly affect custom integrations that call the endpoint directly.

HTTP errors
ResponseCause and fix
400 Invalid Content-Type headerSend Content-Type: application/json with every request.
404 The requested resource does not existThe URL is wrong. Use the endpoint exactly as shown in Getting started, without a trailing slash.
405 Method Not AllowedThe endpoint accepts JSON-RPC messages sent with POST. It does not offer a GET event stream.
406 Not AcceptableSend Accept: application/json, text/event-stream.
429 Too Many RequestsThe public endpoint applies fair-use rate limits. Slow down and retry with exponential backoff. The ratelimit-remaining response header shows how much of your allowance is left.
500 with PayloadParsingErrorThe request body is not valid JSON.
Tool errors

When a tool call fails, the HTTP request still succeeds. The tool result has isError: true, and its text starts with one of the codes below, followed by what went wrong and how to fix it. Most agents can read this message and correct their own request.

CodeCause and fix
INVALID_ARGUMENTSAn argument is missing or invalid, for example a missing locale, a query shorter than 2 characters, a max_results above 50, or a product_code or locale value that is not in the tool schema. Use the exact values from the tool schema and get_available_products.
INVALID_RELEASE_CODEThe release is not available for the chosen product. The error message lists the valid release codes.
RELEASE_CODE_REQUIRES_PRODUCTrelease_code was sent without product_code or product_name. Add a product, or remove the release filter.
INVALID_PRODUCT_SCOPEproduct_code and product_name refer to different products. Send only product_code.
INTERNAL_ERRORThe call could not be completed. First check that the tool name matches tools/list exactly, because an unknown tool name also returns this code. Otherwise it is a temporary server-side problem: retry later, and open an issue if it persists.
No results

If search_help_content returns total_results: 0, remove the release_code, product_code, and sources filters one at a time, or rephrase query using the product's own terms for the feature.

Reporting a problem

Send an x-request-id header with a new UUID on each request. The server returns the same value in its response. Include that value when you open an issue so the request can be traced.

Security and data handling

  • The service is read-only. The tools only retrieve published Help and Support content and cannot change any data.
  • The public endpoint does not require authentication and does not access Autodesk accounts or customer data.
  • Do not put passwords, license keys, access tokens, or other personal or sensitive data in query or in any other tool argument.
  • To report a security vulnerability, follow the Autodesk Vulnerability Disclosure Policy and submit your report through Autodesk on HackerOne. Do not open a public issue for security reports.

About this repository

The ai-integrations repository is the official distribution point for Autodesk AI integrations. This folder, product-help/remote-mcp-server, holds the Autodesk Product Help MCP files.

product-help/remote-mcp-server/
├── server.json   # MCP Registry metadata
├── README.md     # Setup and usage documentation
└── LICENSE.md

server.json follows the MCP Registry schema and is used to publish the server to the Official MCP Registry.

See server.json for the current registry name, version, and remote endpoint.

server.json is maintained by Autodesk and changes only when a new version of the service is published.

Contributing

Contributions to the documentation and client setup examples are welcome.

  1. Fork the repository and create a branch for your change.
  2. Open a pull request that describes the change and how you tested it, for example with the curl check above.

The hosted service is not built from this repository. For feature requests or problems with the service itself, open an issue.

Support

License and terms

LICENSE.md covers the files in this repository only. It does not cover the hosted service or the Autodesk content that the service returns.

Use of the Autodesk Product Help MCP server is subject to the Autodesk Terms of Use and Privacy Statement.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.