Qase Product Updates: July 2026

For a long time, using Qase meant a person in a browser tab. You opened a project, you clicked into a run, you read a dashboard. That's still true, and it's still most of the work. But it's no longer the only way in.

July was mostly about the other doors. We rebuilt our MCP server so an AI agent can drive Qase through the same tasks you'd do by hand, and we made dashboards easier to hand to someone else without rebuilding them or leaving them wide open. July itself was a focused month, so that part is short. I've also used the room to catch up on June at the end, since that update never went out on its own.

Here's what shipped in July.

Pillar 1: MCP, rebuilt around the work you actually do

Qase MCP 2.0. If you've wired an AI assistant into Qase through our MCP server, you know the first version worked but asked a lot of the model. It exposed 83 separate tools. An agent had to load all of them, figure out which one applied, and often guessed wrong. The more tools you give a model, the worse it gets at picking the right one.

We rebuilt the server around that problem. MCP 2.0 collapses those 83 tools into 30 task-oriented ones, organized around what you're actually trying to do instead of every low-level API call.

Two-tier discovery. Instead of loading every tool into context, the agent calls qase_discover_tools to find the ones that fit the task at hand. Less noise for the model to wade through, better tool selection.

Composite workflow tools. Common jobs that used to take a chain of calls now happen in one. Reporting a CI run, triaging a defect, kicking off a regression cycle: each is a single tool call instead of a chain of them.

Full QQL support. Your agent can run QQL queries through MCP, so anything you can ask in the query language, it can ask too.

A REST escape hatch. For anything the task tools don't cover, a qase_api tool calls our REST API directly. You're never boxed in by the curated set.

The server is open source, so you can read exactly what it does before you point an agent at your workspace. In practice, that lets an assistant report results or open defects straight from your workflow, with no hand-copying between windows.

We're also working on a hosted version at mcp.qase.io: a remote instance with one-click sign-in and no local setup. It's not public yet. When it's ready, you won't have to run the server yourself to get started.

Pillar 2: Dashboards worth sharing

Two changes this month, both about getting a dashboard in front of the right people.

Dashboard cloning. Found a dashboard that's almost what you need? You can now clone it in two clicks instead of rebuilding it widget by widget. Pick Clone in the dashboard menu, give it a name, choose the projects, and you're done.

Everything comes along in the copy: every widget and its configuration, including QQL table widgets and all-projects scope. It doesn't have to be your own dashboard, either. Any dashboard you can see can be cloned, so a teammate's is a fine place to start.

The need it answers is specific and common: the same dashboard with a different scope. Same widgets, different projects. Until now the only way to get there was to recreate the whole thing by hand. Cloning turns any good dashboard into a reusable template, and it makes experimenting safe. Clone it, change the copy, and leave the original intact.

Password-protected shared dashboards. Public dashboard links let you share analytics with someone outside Qase, which is handy when the person you want to show doesn't have a seat. The catch was that the link worked for anyone who had it. Share it once and you're trusting it not to travel past the people you meant.

You can now add an optional password to a public dashboard link. Viewers need both the link and the password, so a link on its own reveals nothing. And you can change the password or remove it at any time. It's a small control, and it's what lets you share a dashboard outside your team without worrying about where the link ends up.

A quick catch-up on June

Last month we ran the Q2 wrap-up instead of the usual monthly update, so a few June releases never got a post of their own. The ones worth going back for:

Notion and Microsoft Teams, both two-way. If your requirements live in Notion, they now link straight to test cases, feed Test Designer, and are a source in the traceability matrix, the same depth we already had for Confluence. And Teams became the fourth chat platform we notify into, after Slack, Discord, and Mattermost, so test runs and defects land where your team already talks.

Test Designer started reading your context. The AI that drafts test cases from requirements picked up an "Additional context" field for your own examples and naming conventions, and it can now take screenshots and Markdown docs as input, up to 5 files at a time. Point it at a mockup or a spec and it works from what's actually in front of it.

AI test authoring got rebuilt, and it runs on Android. The generation engine was reworked so you describe a test in plain language and it drafts the steps for you to review, instead of decomposing every action by hand up front. Mobile landed too: upload an Android build, pick a device, and generate tests that run on our cloud. It's an MVP with real edges (no code export, no Flutter yet), but it's live.

Analytics you can ask for in plain English. Natural Language Queries lets you type what you want and hands back the QQL, so you don't have to learn the query language before you get an answer. Dashboard widgets also picked up filtering by milestone, suite, assignee, and more.

There was more under the hood, including a traceability rename to RTM and a round of platform performance work, but those are the pieces worth circling back for. That's the short version, and the reason it's here instead of in its own June post.

Happy testing!