Quant Sentinel launched for AI penetration testing, alongside scoped API tokens, Slack bot integration, AI governance controls, a redesigned dashboard home, Terraform resources for cloud applications and rules, a ScriptAttest REST API, and many Quant Studio improvements.
You can now restrict an API token to specific permission scopes and individual projects, and give it an expiry date. When you create a token, you can pick read-only or full access, combine capability presets (Content Management, Config Management, AI Services, Quant Cloud), or compose a custom scope set. The token list in your profile now shows each token's permissions, project scope and expiry status. The list warns you when a token expires within seven days, and you also receive an email.
New AI platform
Slack bots
You can now deploy AI agents directly into a Slack workspace from the dashboard. A setup wizard steps through workspace connection, agent assignment and access control. Bots support multi-agent routing. When you select several sub-agents, a router agent is created automatically and sends each question to the right sub-agent. Bot configuration includes keyword triggers, custom App Home tab content, and per-user access control with channel restrictions.
A new AI governance page lets you set organisation-wide AI policy. To open it, select AI Governance under AI Services in the dashboard sidebar. You can turn AI services on or off for the organisation, and allow or block specific models. You can enforce a guardrail preset (OFFICIAL, OFFICIAL:Sensitive or PROTECTED) on every AI request. Each preset applies content safety filtering aligned to the Australian Government security classification of the same name. The preset overrides the guardrail settings of agents and requests. You can also set monthly and daily budgets for the organisation and for each user, and the page shows current spend. Governance settings, spend and filter policies are also available through the v3 API.
Compliance skills and customisable managed agents for QuantGov
QuantGov organisations get Quant-managed agents with built-in skills. The skills cover ISM controls, Essential Eight strategies and WCAG 2.2 AA accessibility. An agent loads a skill only when the current task needs it. To change a managed agent for your organisation, open it from the AI agents page and select Customise. You can set its model, temperature and guardrail preset, and add to its system prompt. You can also add or turn off tools and skills, and choose its vector database collections. The base agent itself does not change.
A new Filter Policies section under AI governance lets you define content filtering rules applied to AI input and output. Policies can be assigned as mandatory (enforced across all agents in the organisation) or assigned to individual agents from the agent editor. A built-in test panel lets you validate a policy before saving.
A new GET /api/v3/organizations/{org}/ai/vector-db/collections/{collectionId}/documents endpoint lists the documents in a collection, with limit and offset pagination and an optional key filter to find documents by their key. The Go SDK supports it.
New AI platform
AI chat template
A new AI Chat template deploys a chat interface for your organisation's AI models and agents to Quant Cloud. It streams responses, lets users choose a model or an agent, shows each tool call while it runs, and accepts image and document attachments. Deploy it from the application template list, then add your AI API credentials as environment variables.
Four new GitHub Actions let you manage and test AI agents from a CI/CD pipeline. quant-cloud-agent-deploy-action creates or updates agents from agent definitions and prompt files in your repository. quant-cloud-agent-tools-action registers or updates custom tools. quant-cloud-agent-dispatch-action runs an agent in test mode with tool fixtures and returns its output. quant-cloud-agent-validate-action scores that output against a validation spec, and fails the step if the output does not pass.
New Quant Cloud
Environment deployment history
Each environment has a new Deployments tab. It lists the environment's deployments with their status and who triggered them, including whether a deployment came from the API. You can filter the list to in-progress, completed or failed deployments. The environment details now show when the environment was last deployed. The tab that was named Deployment is now named Build & Push.
You can manage ScriptAttest sites, pages, scans, attestations, sources, policies and schedules through a REST API at https://scriptattest.com/api. Create a key on the API Keys page of the ScriptAttest dashboard and send it in an Authorization: Bearer header. The endpoints are documented at scriptattest.com/docs/api. API keys are available on plans that include API access.
New Sentinel
AI-powered penetration testing
Quant Sentinel is an AI penetration testing tool for web applications and APIs that you are authorised to test. You run it from a Docker image (amd64 and arm64) on your own machine or CI runner. It uses the AI services in your Quant organisation, so you need a Quant API token and organisation ID. To get access to the image, contact Quant sales. Specialist AI agents map the attack surface, analyse each vulnerability class in parallel and attempt a proof-of-concept exploit for each finding. Web application scans cover injection, cross-site scripting (XSS), authentication, authorisation (IDOR) and server-side request forgery (SSRF). API scans add mass assignment, resource abuse, misconfiguration and unsafe API consumption from the OWASP API Security Top 10. Set the target with --target-type (webapp, api or hybrid) and the test mode with --test-mode. White-box mode reads your source code, grey-box mode uses an OpenAPI specification, and black-box mode needs only the URL. A YAML configuration file (--config) sets which paths to focus on or avoid, and how Sentinel signs in. Sign-in options include form login with steps written in plain language, TOTP codes, bearer tokens, API keys and OAuth2. --safe skips exploitation and reports findings as unverified. Each scan writes an executive report and a SARIF 2.1.0 file, and --quiet turns off the terminal dashboard for CI pipelines.
New Sentinel
Infrastructure scanning
Sentinel also assesses network infrastructure as well as applications. Run it with --target-type infra and --targets-file, a file with one IP address, CIDR range, IP range or hostname per line. The scan finds open ports, identifies services and discovers subdomains. Parallel agents then check TLS configuration, network services such as SMB, SNMP, SSH and SMTP, web server configuration and known CVEs. The report lists the web applications that the scan found, so you can test them with an application scan. Infrastructure scans run in black-box or grey-box mode.
New Studio
Brand and identity settings
The Studio settings page now includes a Brand and Identity section for your site name, description, logo, favicon, colour palette (light and dark), fonts, and tone of voice. These settings go into the AI assistant's context so generated content and edge functions match your site's design language. Brand analysis from the migration tool is saved to this section when a migration completes.
A duplicate button is now available on each entry in the collection browser. Duplicated entries are created as drafts with a -copy suffix in the slug and "(Copy)" appended to the title, and the editor opens immediately on the new entry.
New Studio
File move dialogue
You can now move a file to a different path from its right-click context menu in the file browser. A dialogue shows the source and target paths, and you confirm before the move is committed to the current branch in your GitHub repository.
You can now rename edge functions from the function list. A rename option on each function card opens a dialogue to change the filename. Edge function islands embedded in content continue to work after the rename. An island is a <quant-island> tag in an entry. On the page, it shows the HTML that an edge function returns.
New Terraform
Cloud application resources
New Terraform resource types cover the full cloud application stack: quant_application, quant_environment, quant_volume, quant_cron_job, quant_kv_store, and quant_kv_item. You can now manage a complete cloud application in Terraform.
New Terraform
Function, auth, bot challenge, header and serve static rule resources
Provider v5.0.0 adds five rule resources: quant_rule_function, quant_rule_auth, quant_rule_bot_challenge, quant_rule_headers and quant_rule_serve_static. They join the existing redirect, proxy, content filter and custom response rule resources.
The tasks board introduced in February now has status columns that you can drag tasks between, and a sidebar of task lists. You can also drag a task onto a list in the sidebar, and valid columns are highlighted while you drag. A side panel shows each task's details, subtasks, acceptance criteria and dependencies. You cannot manually unblock a task that is blocked by unfinished dependencies.
The agent editor's tool picker now shows platform tools and your custom tools in separate sections, with a shared filter.
Improved Dashboard
Home redesign
The dashboard home has a new two-column layout, with quick actions for your projects beside a summary of your subscription usage. Below them, hits and bandwidth charts can show the last hour, the last 30 days or the last 12 months, with key totals above each chart.
Improved GitHub Actions
Quant Cloud actions run on Node.js 24
The Quant Cloud init, environment, environment sync, environment backup, environment state, environment variables and compose actions now run on Node.js 24 instead of Node.js 20.
Improved Quant Cloud
Additional metrics charts
The cloud application metrics page now includes charts for Network RX/TX, Storage Read/Write IO, and Service Task count, alongside the existing CPU, memory, database, and container charts. Charts with no data are hidden automatically.
Improved Rules
ASN filtering
The rules editor now supports filtering by Autonomous System Number (ASN). You can match or exclude traffic from specific ASNs using an asn_is or asn_is_not condition, complementing the existing IP, country, and traffic source conditions.
Improved Rules
New traffic source options
The traffic source condition can now match the Known abusers and Known crawlers sources, and the Cloud computing services option is now renamed Cloud / hosting providers.
Improved ScriptAttest
Report-only and blocking modes
Each policy version now has an enforcement mode, Report Only or Blocking. In Report Only mode, browsers report violations without blocking anything. In Blocking mode, browsers enforce the policy. New policies start in Report Only, so you can review violations before you switch. Switch modes with the toggle on the CSP Management tab or with POST /api/sites/:id/policies/switch-mode. A switch creates a new policy version with the same directives. Deployments to a linked QuantCDN or QuantGov project now follow the mode: Report Only sends Content-Security-Policy-Report-Only, and Blocking sends Content-Security-Policy. A switch deploys the new header to a linked project automatically, and the policy history shows the mode of each version.
Improved SDKs
API tokens, AI skills and AI governance
The TypeScript SDK shipped v4.11.0 and v4.12.0. Version 4.11.0 adds methods to list, create and delete API tokens. Version 4.12.0 adds methods to list, create, update and delete AI skills, and to import and sync skills and skill collections from GitHub. The Go SDK shipped v4.13.0, v4.14.0, v4.15.0 and v4.15.1, which add vector document listing, AI governance endpoints and the ASN rule condition.
Improved Studio
Async file tree sync with progress
File tree synchronisation from GitHub now runs asynchronously and shows per-batch progress in the UI. The sync skips re-downloading files that have not changed since the last sync, reducing time and API calls for large repositories.
Improved Studio
Unchanged files skipped on publish
When you publish with Quant Builder, pages and assets that have not changed since you last published from your browser are now skipped, unless you select Force. Images and other media now sync on the server instead of passing through your browser.
Improved Studio
Faster publishing with automatic retries
When you publish with Quant Builder, pages now upload in parallel instead of one at a time, and a page upload that fails is retried automatically. Large sites publish faster, and fewer pages fail to publish.
Improved Studio
Preview supports more Astro features
The Studio preview and Quant Builder builds now support content collections that use YAML or JSON data files, <script> blocks that use import statements, and CSS files that use relative @import rules. More existing Astro sites now render as they do in a normal build.
Improved Studio
AI edge function context
When the edge function AI assistant generates a function that returns HTML, it now has access to the project's Tailwind configuration and global CSS. Its system prompt also includes design guidelines that tell it to use the colours and fonts from the Studio Brand and Identity settings. As a result, the HTML that it generates for islands and other parts of a page uses the site's colour scheme, fonts and spacing.
Improved Studio
Locale page routes auto-created on translation
When you create a translation of a content entry, Studio now automatically generates the required Astro page route for that locale if it does not already exist. Translated entries no longer fall through to an incorrect catch-all route.
Improved Studio
Branch indicator
The active branch is now shown as a persistent pill in the status bar at the bottom of the Studio shell. Deploy buttons are now labelled with the current branch name for clarity.
Improved Studio
Persistent active file in URL
Opening a file in the Studio file browser now updates the URL with a ?file= parameter so refreshing the page reopens the same file.
Improved Terraform
ASN condition on rule resources
The redirect, proxy, content filter, custom response, and function rule resources now support an asn condition for filtering traffic by Autonomous System Number, consistent with the dashboard rule editor.
Improved WAF
Log viewer pagination and expandable cells
The WAF log viewer now shows numbered pages, so you can go directly to a page of results. Long URL, domain, and user agent values are shortened by default and expand when you click them, making entries with long paths easier to read.
The payment card input field was showing invisible text in dark mode. The field now uses an explicit dark mode colour scheme.
Fixed AI platform
AI assistant in the tool and edge function editors
The AI assistant in the custom tool editor and the edge function editor could lose earlier messages in a conversation. It could also insert an earlier attempt instead of its final code, or lose the conversation after you deployed a new tool for the first time. The accept and reject controls for suggested code, including in Studio, could fail to appear. These problems are resolved. Also, new AI tools now open on the code tab, and the category field suggests the categories your organisation already uses.
Fixed AI platform
Model list in the workflow editor
The model selector on workflow nodes showed models without their names, so you could not tell them apart or select the one you wanted. Models now show their names and can be selected correctly.
Fixed Studio
Links, branches and media uploads in child environments
When you worked in a child environment, links in the Studio sidebar and navigation rail could open the wrong project. The branch manager could list and create branches in the wrong project, and preview links could point to the wrong address. Media could also be uploaded to the parent project. These now use the correct project and environment.
Fixed Studio
Blank screen after save
The Studio editor could show a blank screen after you saved a file. This is resolved, and if the file tree cannot load, Studio now shows an error that asks you to refresh the page.
Fixed Studio
Cross-collection navigation
Switching between content collections was not refreshing the fields panel, leaving stale fields from the previous collection visible in the entry editor. The panel now updates correctly on cross-collection navigation.