Backup restore across Quant Cloud environments, multilingual search indexing for Drupal and an MCP server for QuantSearch AI, API token secret regeneration, trailing-slash redirects, response time and HTTP status metrics, and reliability improvements across QuantCode.
You can now regenerate an API token's secret from the profile page without deleting and recreating the token. The token keeps its ID, scopes and project assignments, and gets a new secret and a new expiry that you choose. The previous secret stops working, and regenerating an expired token makes it active again. Expired tokens are now shown in a separate section at the bottom of the token list so they remain visible without cluttering the active tokens.
New AI platform
Workflow webhook secrets
Webhook triggers on AI workflows now have reveal and rotate actions on the Triggers page, under AI Services in the dashboard. Reveal shows the current webhook secret, and rotate issues a new one without reconfiguring the workflow.
Projects now have a "Redirect trailing slashes" option in the create and settings forms. When it is on, a request for a path that ends in a slash, such as /about/, gets a permanent (301) redirect to the same path without the slash. The redirect applies to static content that Quant serves, not to requests handled by a proxy rule. New projects, including projects created through the API, have the option on by default, and existing projects keep their current behaviour until you turn it on.
New GitHub Actions
Purge action cache-key and soft-purge inputs (v7)
The purge action has two new inputs: cache_keys (purge by cache key rather than URL pattern) and soft_purge (mark content stale so it is served until fresh content is available). url_pattern is now optional, and cannot be combined with cache_keys.
Completed database and filesystem backups in the dashboard now have a Restore action. You can restore a backup to its own environment or to another environment of the application, after you confirm that the restore overwrites data in the target environment. Restoring to a production environment needs a second confirmation, and a restore in progress shows its status on the backup row. The API also has endpoints to start a restore and to check its status.
New Quant Cloud
Origin protection redirect host
Container settings now include an optional redirect host field for origin protection. When set, requests that fail origin protection checks are redirected to the specified host rather than returning an error.
New QuantSearch AI
Multilingual indexing
The QuantSearch AI Drupal module can now map each language to its own QuantSearch site. When a node is saved, translated or deleted, the module updates the index of the QuantSearch site for that content's language. The Drush commands qs-index, qs-purge and qs-crawl accept a --language flag to target one mapped language or all languages at once. Previously, the module indexed content in all languages into a single QuantSearch site.
New QuantSearch AI
MCP server for AI agents
AI agents and assistants can now set up and manage QuantSearch AI through an MCP server at https://mcp.quantsearch.ai/api/mcp. An agent can create and update sites, start crawls, and push or delete pages directly. It can also run searches, read analytics, and get the embed code for the widget, modal or search page. MCP clients register and sign in automatically with OAuth, using dynamic client registration and PKCE, after you approve access on a consent screen. An agent can even create a QuantSearch AI account for you, so you get working search without leaving the chat. Setup steps for common MCP clients are in the agent guide on quantsearch.ai.
The AI usage page has a new Daily tab with a chart and table covering a rolling 30-day window. A selector lets you view the whole organisation, a single user or a single token. Previously, usage was shown only by calendar month.
Improved CDN
Cache-key purging for QuantGov projects
You can now purge by cache key on QuantGov projects, from the dashboard purge form and the API. A cache-key purge invalidates all content tagged with a shared key in a single request. Soft purge is still not available for QuantGov projects.
The quant purge command no longer requires a URL path argument. You can now run it with only --cache-keys to purge by key without a URL. This is useful in CI pipelines when your origin tags responses with a Cache-Keys header. You cannot combine a path and --cache-keys in one command. This is available from CLI v6.2.0.
The crawler now collects images declared in a page's <head> meta tags, including og:image, og:image:secure_url, twitter:image and schema.org image, and site icons such as favicons and Apple touch icons. Copied sites now keep their social preview images and browser tab icons. Responsive images in pages built with React or Next.js are now also collected.
Improved Dashboard
Page editor revision list sorted and dated
The page editor revision dropdown now shows revisions in descending order (newest first), includes the date each revision was saved, and is capped at 500 entries. Previously revisions were unsorted and showed no dates.
Improved Drupal AI provider
Automatic token refresh
When you connect the Drupal AI provider module to Quant Cloud with OAuth, the module now renews its access token automatically before it expires, so AI features keep working without reconnecting the site.
Improved Drupal AI provider
Longer responses by default
The module's default maximum response length rises from 1,000 to 16,384 tokens, and you can now set it up to 65,536. The default request timeout rises from 30 to 120 seconds, so longer agent and tool-use responses are no longer cut short. When you run the database updates, existing sites that still use the old default values get the new defaults. The updates also raise any configured maximum response length of 4,096 tokens or lower to 16,384.
Run the Drupal database updates (drush updatedb) after you update the module.
Improved Quant Cloud
Response time and HTTP status metrics
The environment metrics page now charts response time and the mix of HTTP status codes (2xx, 4xx and 5xx). These charts replace the network and storage IO charts added in March.
Improved QuantCode
Context overflow handling
Sessions that approach the model's context limit now recover more reliably. The agent correctly identifies overflow conditions, triggers compaction, and resumes rather than stalling. Long-context-capable models use their extended window before compaction is triggered.
Improved QuantCode
Temporary errors retried automatically
Temporary server and connection errors while a response streams, including during context compaction, are now retried automatically a limited number of times instead of ending the turn with an error. QuantCode also now waits up to 120 seconds, up from 45, for the next part of a streamed response before it treats the connection as lost.
Improved QuantCode
Spend counter updates during long-running sessions
The daily and monthly spend figures in the titlebar now refresh periodically while a session is active, not only at the start and end of each turn. Spend is now visible in near-real time during long-running sub-agent work.
Improved QuantCode
Image attachments resized at paste time
Images larger than 500 KB that you paste, drag in or attach in the QuantCode terminal interface are now scaled down and compressed before they are sent to the model. This makes requests smaller and avoids errors from oversized images. Resizing uses the built-in image tools on macOS, or ImageMagick on other systems if installed, and images are sent unchanged when neither is available.
Improved QuantSearch AI
Organisation switcher and invitation links
If you belong to more than one QuantSearch AI organisation, a switcher in the dashboard sidebar now lists them with your role in each. You can move between them without signing out. Invitation links now work when you need to sign in or create an account first. After you do so, the invitation opens and adds you to the organisation. Before, you were sent to the dashboard without joining.
Improved Studio
Markdown preview closer to your Astro site
The live preview now processes Markdown and MDX files more like an Astro build. It wraps .md and .mdx pages in the layout set in their layout frontmatter and applies the remark and rehype plugins from your Astro configuration. In .md files, it also highlights code blocks and loads images that use ~/, @/ or relative paths. Code that reads process.env during preview rendering now gets its values from the .env files in your project.
Improved Studio
Preview URL for the production branch
The Branches screen now shows the preview URL for your production branch, as it already did for environment branches, so all preview URLs are listed in one place.
Improved Terraform
quant_crawler_schedule now supports import
The quant_crawler_schedule resource now implements terraform import, allowing existing crawler schedules to be brought under Terraform management without recreating them.
The expiring-soon badge was being shown on tokens that were not close to expiring. It now correctly appears only on API tokens that expire within the next 7 days.
Fixed AI platform
AI request errors show the reason
When an AI request was rejected, for example because the organisation's monthly AI spend limit was reached, the API returned a generic server error instead of the reason. It now returns the original status and error message.
Fixed Crawler
Crawls with a URL list fetch only those URLs
Crawlers with a configured list of URLs, including their scheduled runs, could follow links from the listed pages and crawl the whole site. These crawls now fetch only the listed URLs and skip link discovery and sitemaps, as single-URL crawls already did.
Fixed Crawler
Finished runs no longer stuck at Provisioning
Some crawler runs stayed at Provisioning in the run list after they had finished. Their status now updates when the run completes.
Fixed Domains & DNS
Zone action buttons
The Delete Zone, Enable CDN, and Disable CDN buttons were not opening their confirmation modal reliably. This is now resolved.
Fixed KV stores
Item actions after search or load more
After you searched the item list or loaded more items, the View, Edit and Delete actions on a KV store item did nothing. They now work after any change to the list.
Fixed Quant Cloud
Metrics chart times in your time zone
Time labels on environment metrics charts, for container and database metrics, were in UTC. They now use your local time zone.
Fixed Rules
Custom error pages for origin errors and timeouts
Custom static error pages on proxy rules were introduced in June 2024. When the origin returned a server error or did not respond in time, these rules did not always show your page. Visitors could see your 404 page or the default error page instead. Quant now shows your custom page for the matching status code, and uses the default error page only when you have not published one.
Fixed Search
Large index updates no longer lose records
When the Drupal module, the WordPress plugin or the search API sent a large batch of records to the search index, the records could be left out of the index. The response still reported them as indexed. Large batches are now indexed reliably, and the response reports any record that cannot be indexed as an error.
Fixed Sentinel
Scans that stopped before the report
Web application and hybrid scans could stop during reconnaissance and end without a report. Long scans could also stop part way through. Both problems are resolved, so these scans now complete with a report.