Troubleshooting
What usually went wrong, and what to try first.
Summarize
Start with the symptom. Most of these point at a setting or a surface, not a reinstall.
Start screen and preview
| Symptom | Try this |
|---|---|
| Preview is empty, or the site will not map | Confirm the host is reachable. Try a sitemap URL, or a list if you already have URLs. Robots or a blocked sitemap can hide the folder map even when the site is up. |
| The path you typed was not found | Silkra falls back to folders it did find. Pick one, or paste a path that exists. |
| Folder count is unknown | The folder exists but was not in the sitemap. Start anyway and use the page limit. |
| Host is unreachable | Edit the URL. Check the host, protocol, and that you can open it in a browser. |
The crawl
| Symptom | Try this |
|---|---|
| Pages are thin or empty | Switch JavaScript rendering to Always render on SPAs. Static HTML only is the wrong call there. |
| Timeouts or a stalled run | Raise request timeout, slow the pace, or lower max pages. Pause, then resume. |
| 403 or a bot wall | Try the Chrome user-agent. Leave robots on unless you run the host. Some hosts you cannot crawl. |
| Rate-limited | The notice shows on the progress view. Conservative pace or Eco mode, then try again. |
| Interrupted banner on the table | Resume, keep the partial version, or delete that version. There is no undo on delete. |
- Stop: keeps what you already have.
- Cancel: tears the run down.
After the crawl
| Symptom | Try this |
|---|---|
| First run is slow, or it is downloading | The first intelligence pass on a machine may download an on-device model, about 35MB, once. |
| Map or semantic search is unavailable | Embeddings and clusters are still running, or this crawl has none. Wait, or check that the run finished. |
| Refresh produced no change log | The first crawl has none. You need a completed refresh (v2 or later). If the refresh failed, there is nothing to compare. |
| Paid features locked after a cold start offline | Sign in again. Last-known plan only holds on a warm session. |
Refreshing a crawl → Change log → Semantic map →
Connectors and MCP
| Symptom | Try this |
|---|---|
| Connector will not connect | Confirm you are signed in and your plan includes integrations. Retry OAuth, or replace the key or credentials requested on the connector page. |
| Card says Reconnect | The saved authorization expired or was revoked. Open the connector and authorize that provider again. |
| Agent says a connected service is unavailable | Open the connector, finish any site, project, board, or team binding, then turn on Available to agent. |
| Search Console columns are missing | Connect Search Console, pick a property that covers this host, and make it available to the workspace. |
| Ahrefs rejects the key | Use an Ahrefs MCP key, not an API v3 key. |
| Data provider returns no result | Confirm the provider plan includes API or MCP access and has units or credits remaining. |
| Webflow cannot match a page | Confirm the bound site owns the crawled URL. CMS and collection item pages are not writable in this version. |
| WordPress reads a page but cannot update SEO fields | Yoast or Rank Math may need their metadata fields exposed to the REST API. |
| Jira, monday.com, or Linear rejects approved work | Confirm the bound destination still exists and check its required fields or board columns. |
| Drive export fails after connecting | Open Google Drive, reconnect the account, then retry the export. |
| MCP client shows no sites | Sync a completed crawl from Manage MCP on the desktop, then reconnect the client. |
Workspaces
| Symptom | Try this |
|---|---|
| Signed out, workspaces still there | Expected. Sign-out does not delete local work. |
| Deleted the wrong workspace | Gone on this device. There is no undo. |
| At the workspace cap | Delete one you do not need, or upgrade. The start screen says so before you scrape. |