> For the complete documentation index, see [llms.txt](https://help.intangible.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.intangible.ai/intangible-mcp-agents/troubleshooting-and-faq.md). # Troubleshooting and FAQ Fixes for connecting an AI agent to Intangible's Model Context Protocol (MCP) server: the server URL, supported clients, sign-in problems, approvals, and empty-scene issues. Three fixes clear many problems connecting an AI agent to Intangible's Model Context Protocol (MCP) server: keep a single Studio tab open, sign in again, or set the Intangible tools to **Always allow**. Each heading below is one symptom, followed by its fix. ## What is the Intangible MCP server URL? `https://mcp.intangible.ai/mcp`. It's a hosted remote server. It uses streamable HTTP and OAuth sign-in with dynamic client registration. **Fix:** Paste the URL into your app's connector or MCP settings. There's nothing to install and no API key. ## Which AI agents can connect to Intangible over MCP? Any MCP client that supports remote servers with OAuth sign-in. **Fix:** Find your app on [Agents & MCP](/intangible-mcp-agents/intangible-mcp-agents.md). Each company has a page with numbered steps. ## Do I need a paid plan to use MCP? You need an Intangible account. Rendering in Visualize uses credits, and other agent work can too, the same as work by hand. See [Credits and tokens](/learn-more/concepts/credits-and-tokens.md). Some agent apps work on their free or personal plans, including ChatGPT Plus and free grok.com. **Fix:** Sign up at [studio.intangible.ai](https://studio.intangible.ai/?utm_source=help.intangible.ai\&utm_medium=docs\&utm_campaign=inline-link\&utm_content=troubleshooting-and-faq), then connect your agent. ## The Intangible viewport stays empty after a prompt Either no Studio tab is open, or the agent is still searching the asset library. On a large scene, the search takes the first couple of minutes. **Fix:** Open a project at [studio.intangible.ai](https://studio.intangible.ai/?utm_source=help.intangible.ai\&utm_medium=docs\&utm_campaign=inline-link\&utm_content=troubleshooting-and-faq), or ask the agent to create one. Then wait. The agent says when it moves from searching to placing. ## The agent reports no open Intangible project or scene The agent builds in the Studio tab open in your browser, and none is open. **Fix:** Ask: *"Create a new project and scene, then start building."* Keep that project open in one Studio tab. ## The agent asks which team to use You belong to more than one Intangible team. The agent needs to know where to create the project. **Fix:** Reply with the team name. ## Several Intangible Studio tabs are open at once With more than one Studio tab open, the agent can edit a scene you aren't looking at. ChatGPT stops and reports an ambiguous target. Gemini can pick the wrong tab and clear that scene before it builds. **Fix:** Close every Studio tab except the project you want. Name that project in your message, then ask the agent to try again. ## The agent asks for approval on every step Your app asks before each tool call by default. A scene build has many steps. **Fix:** Allow the Intangible tools once: * **Claude:** on the Intangible connector page, set the tools to **Always allow**. * **Claude Code:** choose the "don't ask again" option the first time each tool comes up. * **Cursor:** turn on auto-run in Cursor's chat settings. * **Gemini CLI:** choose to allow the tool always when it asks. * **Perplexity and the Gemini app:** both ask before every step. Ask for one zone at a time to keep the approvals short. ## The agent makes a picture instead of building the scene Some apps generate images on their own and pick that over Intangible. **Fix:** Start the prompt with "Use Intangible to…". In the Gemini app, type `@` and pick **Python Mcp Server** first. ## The agent stops halfway through a large scene Apps cap how many steps an agent takes in one turn. Everything placed so far is saved. **Fix:** Reply "continue". For large scenes, pick a model from [Choosing a model](/intangible-mcp-agents/choosing-a-model.md) and ask for one zone at a time. ## Signing in to the Intangible MCP server fails, or a working connection stops The sign-in failed, expired, or went to the wrong Intangible account. **Fix:** Sign in again from your app: * **Claude Code:** run `/mcp`, select **intangible**, and sign in. * **Codex:** run `codex mcp login intangible`, then restart Codex. * **Gemini CLI:** run `/mcp auth intangible`. * **Muse Code:** run `muse mcp login intangible`. * **Claude, ChatGPT, grok.com, Cursor, Perplexity and the Gemini app:** disconnect the Intangible connector and connect it again. On the Intangible sign-in page, check that the page names the app you're connecting. Check that you're signed in to the right Intangible account. ## A connection that worked yesterday stops working Intangible's sign-in expires, and some apps can't renew it on their own. Codex reports that its access token could not be refreshed. **Fix:** Sign in again from your app, as in the section above. ## Codex says the model is not supported with a ChatGPT account Older Codex versions default to a model that ChatGPT sign-ins reject. **Fix:** Update Codex, or pick another model with `/model`. Then send your message again. ## Gemini CLI says "invalid\_grant" or "invalid\_rapt" Your Google Cloud sign-in expired. This only affects Gemini CLI signed in through Vertex AI. **Fix:** Run `gcloud auth application-default login`, then restart Gemini CLI. Your Intangible sign-in is kept. ## The Gemini app has no Custom button for Intangible You're signed in to the Gemini app with a work or school Google account. Google allows custom connected apps, such as Intangible's MCP server, only on personal Google accounts. This is Google's policy. Your Workspace admin can't turn it on. **Fix:** Sign in to gemini.google.com with a personal Google account, or use Gemini CLI, which works with work accounts. See [Google Gemini](/intangible-mcp-agents/google-gemini.md). ## Intangible isn't listed in the Gemini app The Gemini app lists Intangible as **Python Mcp Server**. It doesn't appear under **Active** on the Connectors page, even when it's connected. **Fix:** In a new chat, type `@` and pick **Python Mcp Server**. Don't add Intangible a second time. See [Google Gemini](/intangible-mcp-agents/google-gemini.md). ## Objects land in the wrong place Objects land close to where you asked, and the agent hands the scene back for final positioning. **Fix:** Nudge objects by hand in Build mode. It's faster than another prompt. See [What your agent can do](/intangible-mcp-agents/what-your-agent-can-do.md). ## Does the Intangible MCP server run offline or on localhost? No. The server is hosted at `https://mcp.intangible.ai/mcp`. Your agent reaches it over the network and signs in with OAuth. **Fix:** Use an app that supports remote MCP servers. Apps that only run local servers can't connect. ## How do I remove the Intangible connector? Remove it in the app you added it to. **Fix:** * **Claude Code:** run `claude mcp remove intangible`. * **Codex:** run `codex mcp remove intangible`. * **Cursor, Muse Code and Gemini CLI:** delete the `intangible` entry from the settings file you added it to. * **Claude, ChatGPT, grok.com and Perplexity:** delete the Intangible connector from the connector settings. ## Related * [Agents & MCP](/intangible-mcp-agents/intangible-mcp-agents.md) * [Choosing a model](/intangible-mcp-agents/choosing-a-model.md) * [What your agent can do](/intangible-mcp-agents/what-your-agent-can-do.md) *Last updated 23 October 2026. Behavior may change.* --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://help.intangible.ai/intangible-mcp-agents/troubleshooting-and-faq.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.