Skip to main content
Unstructured Foundation is an Early Access product. Unstructured must accept you into the Early Access program before you can begin using Foundation. Add your name to the Early Access interest list.Unstructured is currently accepting a limited number of participants into the Early Access program at its discretion, with plans to expand access to a more general audience in the future.The following information is being provided to give you an advance preview of possible future functionality. Unstructured makes no express claims as to what extent this information will reflect the state of the product upon initial general availability.
The following sections cover the most common Foundation issues and their causes and solutions. For anything not covered here, request support.

​
Foundation not ready after setup

You set up Foundation and connected to a file source for the first time. When you ask a question, your AI tool says Foundation is not ready or returns an empty result. Possible causes
  • Initial indexing is still in progress. A library of roughly 2,000 files takes about 15 minutes. Larger libraries take proportionally longer.
  • The notification email indicating that Foundation has finished indexing your files has not arrived yet.
Possible solutions
  • Wait for the notification email from Unstructured confirming that Foundation has finished indexing your files.
  • Ask your AI tool directly:
  • If your AI tool still reports that Foundation is indexing your files after the appropriate amount of time (more than 15 minutes per 2,000 files), request support.

​
No notification email after setup

Foundation has finished indexing, but you never received the notification email. Possible causes
  • The email may be in your spam or junk folder.
  • The email address on your account may differ from the inbox you are checking.
Possible solutions
  • Check your spam or junk folder for a message from Unstructured.
  • If your AI tool confirms Foundation is ready, your index is available. No action is required.

​
Login access code missing or rejected

You entered your email address on the Foundation login page, but no access code arrived, or the code you entered was rejected as invalid. Possible causes
  • The access code arrives in its own email, sent after you submit your email address. Your Early Access acceptance email does not contain a code.
  • Starting authentication again sends a new access code and invalidates earlier codes. Repeated attempts leave several codes in your inbox.
  • The Foundation login page rejects your email address if your account is not yet accepted into the Early Access program.
Possible solutions
  • Look for a second email from Unstructured, sent after you submitted your email address. This is the email that contains your access code.
  • If you started authenticating more than once, enter the code from the most recent email. Earlier codes no longer work.
  • If repeated attempts left several browser tabs open, close them, start authenticating again, and use the code from the newest email.
  • If the login page rejects your email address, confirm that you have received an Early Access acceptance email from Unstructured. If you have not, add your name to the Early Access interest list.

​
Connector requires re-authorization

Your AI tool or a notification email tells you that a connector needs to be re-authorized. A connector may also appear in an error or warning state when you ask your AI tool about your index. Possible causes
  • OAuth tokens have a fixed expiry period. For example, Google OAuth tokens expire every seven days. Other OAuth providers also expire on a schedule.
  • The account that authorized the connector has had its access revoked or permissions changed in the source system.
Possible solutions
  • Click the re-authorization link in the AI tool message or notification email. This completes the OAuth flow without leaving your AI tool.
  • After re-authorization, incremental indexing resumes automatically. Files from the connector stay searchable from the last successful index while re-authorization is pending.

​
No results returned by queries

You ask your AI tool a question about your files and it says it could not find anything relevant. Possible causes
  • A connector may be in an error or stale state, so its files are not in the current index.
  • Initial indexing may not yet be complete.
  • The query may be too specific or use terminology not present in the indexed files.
Possible solutions
  • Ask your AI tool about the state of your index:
    Your AI tool returns a list of connected sources, file counts, and the timestamp of the last successful index for each. Connectors in an error or stale state appear here.
  • Re-authorize any connector shown in an error or stale state. See Connector requires re-authorization.
  • Try a broader or differently phrased question. Foundation retrieves files based on topics, entities, summaries, and full text, not exact keyword matching.

​
Foundation temporarily unavailable

Your AI tool reports that Foundation is temporarily unavailable. Questions about files go unanswered, but other tasks work normally. Possible causes
  • Unstructured is experiencing a temporary service disruption.
Possible solutions
  • Your AI tool continues to work normally for non-Foundation tasks. Only file queries are affected.
  • Try again after a few minutes. When Foundation comes back online, your AI tool resumes without reconfiguration.
  • If the issue persists beyond an hour, request support.

​
New chat answers from the web instead of Foundation

You start a new chat and ask about your files. Your AI tool answers from the web or from its general knowledge rather than from your indexed files. Possible causes
  • Each new chat starts without Foundation context. Your AI tool does not carry over Foundation from an earlier chat.
  • The request does not name Foundation, so your AI tool routed it to another tool such as web search.
Possible solutions
  • Name Foundation in the request, for example Use Foundation to summarize what changed in our pricing docs last quarter.
  • Confirm that your index is reachable by asking your AI tool:
  • To make the routing consistent across chats, add a standing instruction to your AI tool’s rules or memory file, such as CLAUDE.md for Claude Code.
For phrasing guidance, see When and where the server is used.

​
AI tool not connecting to Foundation

Your AI tool does not appear to be using Foundation. Questions about your files go unanswered, or your AI tool says it has no access to your files. Possible causes
  • Your request may not name Foundation, so your AI tool routed it elsewhere, such as to web search. This is the most likely cause if Foundation worked previously.
  • Your AI tool’s configuration file may be missing the Foundation MCP server entry.
  • Claude Desktop or other AI tools may need a restart after the MCP configuration is added.
Possible solutions
  • If Foundation worked previously, name Foundation in the request before changing any configuration. See New chat answers from the web instead of Foundation.
  • For Claude Code: Check your .mcp.json file for the Foundation MCP server entry. If it is missing, try re-installing the Foundation MCP server.
  • For Claude Desktop and other AI tools: Confirm that the Foundation MCP server is listed in your AI tool’s configuration file.
  • Restart your AI tool after any changes to its configuration file.
For setup details, see The Foundation MCP server.

​
Next steps

​
Additional resources

For help with any of the issues on this page, or for issues not listed here, request support.