All months

Changelog · 2026-06

June 2026

Managed cache for cloud applications, OpenAI-compatible AI endpoints, per-interface AI spend caps and per-user overrides, QuantCode document attachments and supply-chain reviews, and a stable release of the Drupal module for Drupal 11 sites.

30 updates across 13 products shipped this month.

New 5

Back to top
New API

Subscription usage endpoints

A new endpoint, GET /api/v3/organizations/{org}/subscription/cloud-usage, returns the per-project compute, database and storage costs shown on the subscription page. It covers the requested month and the month before it. GET /api/v3/organizations/{org}/ai/usage now returns the same AI usage figures as the subscription page, grouped by model, user or token, with an optional 30-day daily series. Both endpoints need the new subscription:read token scope, and the token owner must have permission to manage the organisation's subscription. This is a breaking change for /ai/usage: it previously needed AI access and the ai:use scope, and its response format has changed. /ai/usage/me is unchanged.

Update integrations that call /ai/usage to use a token with the subscription:read scope and to read the new response format.

New Quant Cloud

Managed cache for applications

Cloud applications can now include a managed cache. It is a Redis-compatible (Valkey) cache that all environments of the application share. You can enable it with a configurable storage cap when you create an application from the manual configuration form. For an existing application, you can enable or remove it on the application's Environments page. Provisioning runs in the background. To connect an environment, edit and save it: this adds the cache connection details as environment variables and redeploys the environment.

New QuantCode

Document attachments

You can now attach Word and Excel documents (.docx, .xlsx, .doc, .xls) to conversations in the QuantCode web app, desktop app, and terminal interface. Documents are sent to the model alongside images and PDFs.

New QuantCode

Supply-chain review agent

A new supply-chain agent checks third-party dependencies in autonomous jobs. When it reviews a pull request that updates dependencies, it checks only the dependencies that changed, including any new packages that the lockfile update brings in. It looks for known vulnerabilities and signs of a compromised package, flags any change that is not a dependency change, and posts its findings as a review comment. It can also audit a whole repository and record the results in a GitHub issue, or open a pull request that pins dependencies to exact versions and GitHub Actions to specific commits.

Related documentation: Supply-chain agent

New Sentinel

Evidence capture through an intercepting proxy

You can now route Sentinel's web traffic to the target through an intercepting proxy, such as Burp Suite. The proxy then holds a full, replayable record of each request and response. Set --proxy <url> or the proxy block in the YAML configuration. The proxy receives traffic from the test browser, the HTTP and GraphQL tools and the command-line tools that agents run. Infrastructure port scans do not go through the proxy. To capture HTTPS with certificate checks on, add the proxy's CA certificate with --proxy-ca. Sentinel's own requests to Quant do not go through the proxy, so your Quant API token is not captured.

Improved 16

Back to top
Improved Account

Free trial starts with a payment card

The 30-day free trial introduced in February now starts when your organisation first creates a resource, such as a project or an application. You choose a plan and add a payment card, and you are not charged for 30 days. Trials that started without a payment method also need one before you create more resources. If you register or sign in from a team invitation, you join the inviting organisation automatically.

Improved AI platform

Dedicated OpenAI-compatible endpoints

After January's OpenAI-compatible request format, the AI platform now has dedicated /oai/v1 endpoints for chat completions (buffered and streaming), embeddings and model listing. Quant identifies the organisation from the bearer token, so you can set an OpenAI SDK's base URL to https://dashboard.quantcdn.io/oai/v1 with no other changes. On QuantGov, use https://dash.quantgov.cloud/oai/v1.

Related documentation: OpenAI compatibility

Improved AI platform

Per-interface spend caps and per-user overrides

Building on the existing per-user daily and monthly budgets, organisation admins can now set total daily and monthly spend caps for each AI interface in AI governance. An AI interface is the channel that requests come through, such as Slack, the API gateway, streaming, workflows, orchestrations or embeddings. Admins can also override the per-user budget for individual users, with an Unlimited option that exempts a user from it. Spend from autonomous QuantCode jobs is now tracked under its own identity, system:code-agent. It no longer counts towards the budget of the user who owns the job's token, and you can give it a separate budget with an override. To help you find user IDs, the AI usage breakdown now shows each user's ID, and the override field suggests known system identities.

Related documentation: AI Governance: Spend Limits

Improved AI platform

Workflow editor drafts and publishing

The workflow editor now keeps your edits in a draft, separate from the published workflow. You can change a workflow without affecting live runs, and then publish the draft when it is ready. Published workflows now open read-only, with a prompt to edit them as a draft. For tools that start an asynchronous job, a new Await result option makes the workflow wait for the job to finish, so later steps receive the job's completed result. An awaits badge marks those steps on the workflow canvas. The configuration panel and the condition labels on the connections between steps are also easier to read.

Improved Cloud CLI

--config flag for visual regression testing

From Cloud CLI v0.5.0, qc vrt accepts a --config <path> flag, so you can use a specific visual regression testing config file instead of ~/.quant/vrt-config.json, for example one generated by your infrastructure tooling. If the file given with --config is missing or invalid, the command stops with an error instead of falling back to the default file.

Improved Drupal module

Stable release for Drupal 11

Version 2.0.0 of the Quant Drupal module is now a stable release for Drupal 11, and 1.7.0 is the release for Drupal 10. Drupal 9 is no longer supported: 1.7.0 requires Drupal 10.

Related documentation: Get started with Quant and Drupal

Improved Drupal module

Translated home pages

On sites that use language path prefixes, the module now pushes the home page at each language prefix, such as /en or /fr, to Quant. Before, it pushed them only when the front page was a node you saved or seeded. It now pushes them with the default home page and the error pages when you run a seed or save the Basic site settings form. The change is in versions 1.7.0 (Drupal 10) and 2.0.0 (Drupal 11).

Improved Drupal module

Deleted content handling

Versions 1.7.0 (Drupal 10) and 2.0.0 (Drupal 11) unpublish more content on Quant when you delete it in Drupal. Deleting a node or taxonomy term now also unpublishes its translations. Deleting one translation of a taxonomy term now unpublishes that translation's page, as it already did for nodes. Deleting an audio, document, image or video media item unpublishes its file. Deleting an image file or image media item also unpublishes its image style files. Deleting a Quant Search page now unpublishes its page.

Improved Drupal module

Queue runs for several projects on one server

drush quant:run-queue now keeps a separate lock for each Quant project. Sites on one server that push to different projects no longer block each other's queue runs. When a run is already in progress, the message names the lock file and suggests drush quant:unlock-queue. The change is in versions 1.7.0 (Drupal 10) and 2.0.0 (Drupal 11).

Improved Drupal module

Warnings for pages that need static-site setup

Version 2.0.0 (Drupal 11) warns administrators about content that needs extra setup on a static site. A warning appears on pages with a view that has exposed filters, and on nodes whose Layout Builder layout contains a webform. Each warning links to the Drupal integration documentation. Version 1.7.0 for Drupal 10 does not show these warnings.

Improved QuantCode

More accurate autonomous reviews

In autonomous jobs, the code review, security review, debug and supply-chain agents can now look up where code is defined and called across the repository, instead of judging a change from the diff alone. Before a review is posted, a second pass checks each finding against the source code and drops findings it cannot confirm.

Improved SDKs

TypeScript and Go clients v4.19.0

The TypeScript and Go clients are updated to API v4.19.0. New coverage includes the AI Search API, database and filesystem restores, the subscription cloud-usage endpoint, and the new AI governance spend-limit fields. AI usage statistics (getAIUsageStats) move from the AI Monitoring API to the AI API and return the new /ai/usage response. The scaling policy methods take new parameters, and get and update are renamed to list and upsert (for example, listScalingPolicies and upsertScalingPolicy in TypeScript).

Update code that calls getAIUsageStats or the scaling policy methods when you upgrade to v4.19.0.

Improved Search

AI search embed and upgrade options

You can now embed AI search as a modal search dialogue that opens from a search button or with Cmd/Ctrl+K. You can also set the display URL used in result links from the Result Links settings. Projects on traditional search gain an Upgrade to AI Search action on the search dashboard. Upgrading makes AI search the project's only search mode and clears the existing keyword index and its extractor configuration. You cannot undo this, so the dashboard asks you to confirm first. Projects stay on traditional search until you upgrade.

Improved Sentinel

Signed-in sessions, new checks and finding status

Sentinel cannot automate some logins, such as some single sign-on flows, CAPTCHAs and SMS or push multi-factor authentication. For these, you can now give it a session that is already signed in (login_type: session in the YAML configuration). Web application and hybrid scans add a cross-site request forgery (CSRF) track. White-box scans also check your source code for hard-coded secrets and your dependencies for packages with known vulnerabilities. Infrastructure scans now re-test each finding, and add UDP port scanning, SNMPv3 checks and subdomain takeover detection. --aggressive-validate allows intrusive re-test checks, and --safe skips the re-test. Reports start with a findings summary that marks each finding Confirmed or Unverified, and SARIF results include the file and line of each finding.

Improved Studio

Tailwind 4 support

The Studio AI assistant now detects whether your project uses Tailwind 3 or Tailwind 4. It updates theme colours and fonts in the right place: the @theme block in your global stylesheet for Tailwind 4, or your tailwind.config file for Tailwind 3. Sites that the site migration tool builds now use Tailwind 4 too.

Improved Terraform

AI governance spend caps and redirect host support

The quant_ai_governance resource now supports per-interface spend limits and per-user overrides, and container origin protection configuration accepts the new redirect_host field. The provider has been updated to the v4.19.0 API client.

Fixed 9

Back to top
Fixed AI platform

GitHub webhook errors on workflows

GitHub webhook events sent to some workflows created or updated from the dashboard returned a "webhook not configured" error. This is now resolved: new and updated workflows register correctly when you save them.

Save again any existing workflow that still returns a "webhook not configured" error.

Fixed Crawler

One failing schedule no longer stops the others

An error in one crawler schedule could stop other crawler schedules from running. Each schedule now runs on its own, so an error now skips only that one schedule's run.

Fixed Drupal module

Missing CSS and JavaScript on pushed pages

On Drupal 10.1 and later, some pages reached Quant without their aggregated CSS or JavaScript files and showed unstyled. Versions 1.7.0 (Drupal 10) and 2.0.0 (Drupal 11) now push these files under the file name that each page uses.

After you update the module, run a seed again so that any missing CSS and JavaScript files are pushed.

Fixed Drupal module

Search page query parameter errors

Quant Search pages could produce unknown query parameter errors. Versions 1.7.0 and 2.0.0 fix this: search pages no longer update the URL as you search. A link with a keys search term still opens the search page with that term. Later searches and filters are not kept.

Fixed Drupal module

Remote video embeds on the static site

Remote videos, such as YouTube videos added through Drupal's Remote video media type, could fail to display on the static site. Versions 1.7.0 (Drupal 10) and 2.0.0 (Drupal 11) rewrite the video embed URL so that it loads from your static site. This now works even when Strip host domain from content is turned off.

Fixed QuantCode

PDF attachments read reliably

The model could fail to read a PDF that you attached to a conversation. This is now resolved.

Fixed QuantCode

Cleaner response rendering

The web and desktop renderers now remove internal formatting tags from model responses, so output renders cleanly with all models.

Fixed QuantCode

Web search and fetch tools return results

The web search, web page fetch and web browser tools could hang or repeat without returning a result, which stalled multi-step tasks that depended on them. These tools now return their results, and multi-step tasks continue as expected.

Fixed QuantCode

Autonomous jobs recover and report correctly

Resuming an earlier autonomous job could fail. Resumed jobs now continue on the earlier job's working branch, or on a new branch when the earlier job did not push one. Jobs no longer stall until they time out when a step asks a question that nobody is there to answer, and pull request links in job results now display correctly.