> ## Documentation index
> Fetch the complete documentation index at: [llms.txt](../llms.txt)
> Use this file to discover all available pages before exploring further.

# Errors and limits

Every limit an agent can hit on the Sleuth MCP server, and what to do about each error.

## Limits

| What | Limit |
| --- | --- |
| File upload through MCP | 10 MB per file, decoded. Upload larger files in Studio. |
| Rows per `get_dataset` call | 50 by default, up to 2,000 |
| Rows per dataset in `export_project` | 10,000 |
| One `wait_for_update` call | Up to 300 seconds. Chain calls for longer runs. |
| Messages per `get_messages` call | Up to 200 |
| Projects per `list_projects` call | Up to 100 |
| Results per research library search | Up to 100 |
| Passages per `query_research_document` call | Up to 50 |
| Matches per `search_documents` call | Up to 50 |
| Download links | Short-lived. Ask for a new one. |
| Research runs at once, per API key | 10. More returns `run_limit_exceeded`. |
| Explore datasets | Row caps by anchor. `explore_get_filters` lists them. |

## Errors and fixes

| You see | What it means | Fix |
| --- | --- | --- |
| `onboarding_required` | The account isn't set up yet. | Call `get_onboarding_status` and follow `next_action`. [Onboarding](tools-onboarding.md) |
| Waiting for approval | A new access request is with Sleuth. | Tell the user they'll get an email, then reconnect. |
| Out of credits, trial ended, or payment failed | The workspace can't start new research. Sleuth's message says which, and links to plans or billing. | Show Sleuth's message as it is. [If you run out of credits](out-of-credits.md) |
| `run_limit_exceeded` | An API key already has 10 research runs going. | Wait for one to finish, then send again. |
| `409` or "project busy" | A research run is already going on that project. | Don't send again. Wait with `wait_for_update` until the run finishes, or send `cancel` if the user wants to stop it. |
| "Can't reach the storage host" when fetching a link | The AI tool's sandbox can't reach the download address. That's the client, not Sleuth. | Give the user the link to open in a browser. |
| `403` with an API key | API access isn't on for this organization. | Contact Sleuth. [API](api.md) |
| `422` from `get_artifact_svg` | The artifact isn't a chart. | Use `get_artifact` instead. |
| No matches right after an upload | The document is still being indexed. | Wait a minute and search again. |
| Ambiguous filter value | A term matched more than one Explore value. | Ask the user which one they meant. |
| Expired or broken download link | Links last minutes. | Call the tool again for a fresh link. |
| Sell-side download refused | The organization doesn't have download access. | Read passages with `query_research_document`, or talk to Sleuth about access. |
| Run status `failed` | The run stopped. It doesn't use credits. | Read the last message, then try again or contact Sleuth. |
| Sign-in loops or "not allowed" | The client didn't finish OAuth, or the connection is stale. | Remove and re-add the connector, then sign in again. [Fix connection problems](ai-troubleshooting.md) |

For anything else, the user can email support@sleuthinsights.com.

## Related

- [Fix connection problems](ai-troubleshooting.md): The most common problems connecting Sleuth to an AI tool, and how to fix each one.
- [How a research run works](mcp-run-lifecycle.md): Create a project, send the question, wait for the status to change, handle approvals and clarifications, then read the datasets and the answer.
- [Get help](get-help.md): Where to ask, depending on what you need.

*Updated November 9, 2026.*
