All months

Changelog · 2026-04

April 2026

QuantCode launches as an AI coding agent for the terminal and desktop, with autonomous jobs that open pull requests; other changes include a Studio live preview that runs your Astro configuration, a new AI Search v3 API and Terraform resources for AI services.

37 updates across 13 products shipped this month.

New 17

Back to top
New AI platform

Model picker for governance-approved models

The tool testing panel and the Studio AI assistant now show a model selector. It lists the models that your organisation's AI governance settings allow and marks the currently recommended model as the default.

New Quant Cloud

EmDash CMS application template

You can now create an EmDash CMS application on Quant Cloud from a template. EmDash is an open-source TypeScript CMS built on Astro. By default it keeps content in a SQLite database on a persistent volume. It switches to PostgreSQL automatically when you add a managed database to the application. To start, use the Deploy to Quant Cloud button in the template's README, or choose EmDash when you create an application from a template. Starter kit deployments now also generate the secrets that a template declares, such as signing keys, so the application starts without manual setup.

New QuantCode

Terminal coding agent launch

QuantCode is an AI coding agent that runs in your terminal and gets its models only from your organisation's Quant AI services. The first release, 1.0.0, is for Macs with Apple Silicon, and you start it with quantcode or qcode. To install it, run npm install -g @quantcdn/quantcode. To set it up, choose QuantCDN or QuantGov, sign in with your Quant account, and select an organisation. You choose from the models that your organisation's AI governance settings allow, and the default is Claude Sonnet 4.6. QuantCode reads and edits files, runs shell commands and gives tasks to sub-agents. It formats the files it edits and shows language server errors. You can add MCP servers, skills, custom slash commands and AGENTS.md project instructions, and you can resume, fork and undo sessions.

New QuantCode

Organisation agents and handoffs

Agents that your organisation creates in the Quant dashboard appear in QuantCode with their system prompts, models and skills. You switch agents with Tab. QuantGov organisations also get code, plan, security, review and debug agents that Quant manages, and the code agent is the default. When an agent completes its part of a task, it can suggest a handoff to one of these managed agents, and you confirm the switch. When you change organisation with /org, QuantCode loads the models and agents of that organisation.

New QuantCode

Desktop app and Homebrew install

QuantCode now has a desktop app for Macs with Apple Silicon. You sign in with your Quant account and use the same models and agents as in the terminal interface. To install the terminal interface, run brew install quantcdn/tap/quantcode. To install the desktop app, run brew install --cask quantcdn/tap/quantcode-desktop. You can also download both from the QuantCode releases page on GitHub (quantcdn/quantcode-releases). The terminal interface also runs on Linux (x64 and arm64). New versions are available only from Homebrew and the releases page. The npm package stays at version 1.0.0.

New QuantCode

Autonomous coding jobs on GitHub repositories

QuantCode can now work on your GitHub repositories as an autonomous job, without user interaction. To set it up, open the new QuantCode Agent page under AI Services in the dashboard. Install the QuantCode Agent GitHub App on your GitHub organisations or account, and choose the repositories. An AI platform Slack bot or workflow can start a job with one or more tasks. For a change request, the job plans the work, writes the code and tests, reviews the result and opens a pull request. For a review request, the job posts the review on the pull request. The QuantCode Agent page lists your organisation's jobs with their status, repository and a link to the pull request.

New QuantCode

AI spend display

QuantCode shows your AI spend for today and for this month. If your organisation sets per-user daily or monthly budgets, QuantCode also shows each budget and the percentage that you used. The spend is in the sidebar of the terminal interface and in the title bar of the desktop app. When you have used 80% or more of your daily budget, the desktop app shows a warning when you send a message.

New QuantSearch AI

Drupal faceted search and metadata field mapping

The quantsearch_ai Drupal module now supports flexible metadata field mapping. You can map any Drupal field to a custom metadata key that is sent to the search index, instead of relying on fixed title, body, summary and tags fields. You can enable faceted search on the search page block, and configure preset filters (JSON) on each search widget block. A preset filter applies to every search from that block, for example {"content_type":"policy"}. The block configuration form validates preset filter JSON before it saves.

New QuantSearch AI

Team members and invitations

On Enterprise plans, you can now invite people to your QuantSearch AI organisation from the dashboard settings page. You give each invitation the admin or member role. Invitees accept from the link in the email they receive. From the same page you can see the organisation's members, change their roles and remove them.

New Search

AI search v3 API

A new project-scoped AI Search API is available under /api/v3/organizations/{org}/projects/{project}/ai-search/. It is the first documented API for AI Search in Quant projects. It does not apply to the separate QuantSearch AI service at quantsearch.ai. It covers the full lifecycle. You can enable or disable AI Search and manage its settings (public access, request limits). You can ingest pages, list indexed pages with cursor pagination, and delete the index. You can also run semantic search and AI chat, get usage statistics and top query analytics, and trigger a crawl. The API documentation now covers these endpoints.

Related documentation: Managing your search index: API

New Studio

Rich-text editing in the preview

When you edit an entry, you can right-click text in the live preview to make it bold, italic, underlined, struck through, code or a link. The same menu changes a block to a paragraph or heading, and inserts a paragraph, heading, list, code block or image below it. Double-click an image in the body to replace it from your media.

New Studio

Component props panel

Clicking a component in the live preview opens a resizable props panel with that instance's values. Changes you make there re-render the preview straight away, so you can try values out. They are not saved to the page source.

New Studio

Unsaved edits kept across page reloads

The Studio file editor now stores unsaved edits in browser local storage and restores them automatically if you refresh or navigate away before saving. Unsaved edits are stored per project and branch, so work on different branches stays separate. A dot marks files with unsaved changes, and a "Discard changes" action lets you abandon those edits when needed.

New Terraform

AI agent and Slack bot resources

The Terraform provider now includes quant_ai_agent and quant_slack_bot resources, allowing AI agents and their associated Slack bots to be managed as infrastructure. Both resources support full CRUD and import.

New Terraform

AI skill, vector database and governance resources

Provider v5.2.0 adds four resources: quant_ai_skill manages AI skills that you define inline, quant_ai_vector_collection and quant_ai_vector_document manage vector database collections and their documents, and quant_ai_governance manages your organisation's AI governance settings as code.

New Terraform

AI custom tool resource

A new quant_ai_custom_tool resource lets you define custom AI tools in Terraform. You can supply JavaScript edge function code directly in the resource configuration — the platform deploys it automatically and registers the tool endpoint.

New Terraform

AI agent overlay resource

A new quant_ai_agent_overlay resource lets your organisation customise a global AI agent without changing the base agent. A global agent is an agent that Quant manages, such as the QuantGov managed agent quantgov-code. You set its ID in agent_id, and the agents API marks global agents with isGlobal. An overlay can set the model and temperature, add text to the system prompt, add or disable tools and skills, choose vector collections and set a guardrail preset.

Improved 11

Back to top
Improved AI platform

Slack bot API manages the bot's agent

The Slack bot API now creates and updates each bot's AI agent for you. Creating a bot now requires a name, system prompt and model, not the ID of an existing agent. When you create or update a bot, you can send settings such as the temperature, tools and guardrail preset. Bot API responses include these settings, so you no longer need a separate agent lookup. Deleting a bot also deletes its agent.

Update Slack bot create requests to send a name, system prompt and model instead of the ID of an existing agent.

Related documentation: Configuration reference: API management

Improved AI platform

Custom tool API deploys edge function code

Creating a custom tool through the API (POST /api/v3/organizations/{org}/ai/custom-tools) now requires the tool's JavaScript source in edgeFunctionCode and no longer accepts edgeFunctionUrl. In one request, the platform deploys the edge function, sets the tool's URL and registers the tool. If you send a create request with the name of an existing tool, the platform updates that tool and redeploys its code. Reading a tool returns its deployed code, and deleting a tool also deletes its edge function. A new GET /api/v3/organizations/{org}/ai/custom-tools/{toolName} endpoint returns a single tool.

Send the tool's JavaScript source in edgeFunctionCode, not edgeFunctionUrl, when creating a custom tool through the API.

Related documentation: Creating AI tools: Tool components

Improved AI platform

Slack bot group DM support

Slack bots can now take part in multi-person direct message conversations (group DMs). The bot's Slack app configuration now includes the required mpim scopes and message event subscriptions.

Select Reinstall to Slack in the settings of each bot that you connected before this change, to grant the new scopes.

Improved AI platform

Privacy and Quant platform skills for managed agents

Quant-managed agents on QuantGov have two more built-in skills: the Australian Privacy Principles and the Quant platform. As with the other skills, an agent loads them only when the current task needs them.

Improved AI platform

Tasks board dependency graph

The tasks board now has a graph view that shows how the tasks in the selected list depend on each other. You can pan and zoom the graph, and fit it to the view.

Related documentation: Task management: Dependencies

Improved Cloud CLI

New documentation section

The documentation has a new Cloud CLI section covering installation and sign-in, a command reference, SSH access, backups and visual regression testing. The Quant Cloud documentation also has new pages on managing backups and syncing environments, and now describes all four ways to create an application.

Related documentation: Cloud CLI

Improved QuantCode

Longer task descriptions for autonomous jobs

Each task in an autonomous coding job can now be up to 20,000 characters, up from 2,000, so you no longer need to split detailed multi-file instructions across tasks. A job that does not name a base branch now starts from the repository's default branch, not always from main.

Improved SDKs

Slack bots, AI governance and custom tools

The TypeScript SDK shipped v4.15.4, which adds Slack bots, AI governance, filter policies, agent overlays and the ASN rule condition. The Go SDK shipped v4.15.3 to v4.15.8, which add a document key to vector document uploads, the Slack bot API's inline agent fields and custom tools with inline edge function code.

Improved Studio

Live preview runs your Astro configuration

The live preview now evaluates your site's astro.config file and runs its Astro integrations in the browser, so pages that depend on integration settings render as they do on your site. The preview scales a desktop-width page to fit the panel and loads assets referenced with ~/ paths. When an AI assistant edit causes new errors in the preview, Studio sends the errors back to the assistant to fix. Studio does this up to three times, then shows any remaining errors in the chat.

Improved Studio

Unpublished changes banner and file change badges

The file browser now shows a banner with the number of unpublished changes on your working branch, building on the editorial workflow introduced in February. The file tree marks each changed file as modified, added, deleted or renamed (M, A, D or R), and the collection browser marks new and deleted entries. The banner's Review button opens the branch's open pull request, or creates one if none exists.

Improved Studio

Content-friendly pull request diffs and file revert

The pull request page now shows changes to an entry field by field, with the body side by side, and shows a line-by-line diff for code files. You can also revert a changed file or entry to its production branch state from the file tree or collection browser, after you confirm.

Fixed 9

Back to top
Fixed Cloud CLI

Environment logs

qc env logs returned no entries. Version 0.4.3 fixes this, and the project and crawler pickers in crawler commands filter as you type.

Update the Cloud CLI to version 0.4.3 or later.

Fixed Crawler

File uploads and page limit on large crawls

On large crawls, some files found on pages, such as images, were not uploaded to the project's static copy before the crawl ended. Link discovery could also stop before the crawl reached its max_html page limit. Crawls now continue to process files as long as downloads make progress, and continue to discover pages until they reach the limit. File downloads no longer follow the page delay setting and run at up to 300 requests per minute, so a long page delay no longer slows them.

Fixed Dashboard

GitHub linking without a Quant Cloud subscription

Users without a Quant Cloud subscription could not link a GitHub account, which Studio needs. They now can, from the dashboard.

Fixed KV stores

Link and unlink from a project

The actions to link a KV store to a project, and to unlink it, were missing from the KV store list. They are available again: Link opens a project picker and Unlink asks you to confirm. The global KV stores page also failed to load with no project selected, and now opens without one.

Fixed Projects

Manual project creation without site search

Creating a project manually failed with "No valid search subscription found" for organisations without a search subscription, even when search was not turned on. You can now create the project without search.

Fixed Studio

Projects with a production branch other than main

In a repository whose default branch is not main, such as develop, Studio could show one branch while it read and saved files on another. Studio now detects the default branch when you import a repository and uses it for branch comparisons, reverts, review pull requests and deployments. You can change it in Studio settings under Production branch.

Fixed Studio

Entry action buttons and file delete

Clicking delete, duplicate or translate on a collection entry also opened the entry in the editor behind the confirmation dialogue; it no longer does. Deleting a file that existed only on a working branch could also fail with an error — file deletes now work on any branch.

Fixed Studio

Media picker opens as full-viewport modal

The media picker was opening confined to the sidebar instead of covering the viewport. It now opens as a full-viewport modal again, consistent with its original design.

Fixed Terraform

Unexpected changes in AI resource plans

Plans for the new quant_ai_governance, AI agent, agent overlay, Slack bot and custom tool resources could show changes after a successful apply, for example to temperature values or a custom tool's timeout. Provider v5.10.4 and later resolve these differences, and you can now clear a guardrail preset by setting guardrail_preset to an empty string.

Update the Terraform provider to v5.10.4 or later.