Google Gemini has quietly shipped a wave of updates that seriously change what the platform can do. Some of them look small at first glance, but they are actually a huge deal once you understand the implications. The biggest one is custom MCP support inside Gemini Spark, which opens the door to connecting Gemini to almost any app you want. On top of that, Spark scheduling is much more advanced, Notebook can now generate short form video content, and AI Studio keeps getting more powerful with new image models, app building tools, and agent creation.
If you use Gemini for automation, content creation, app development, or AI workflows, these updates matter. A lot.
Why these hidden Gemini updates matter
What makes these releases interesting is not just that new buttons appeared in the interface. It is that Google is steadily turning Gemini into a much broader operating layer for work, content, and automation.
There are really four major changes here:
- Gemini Spark can now connect to custom apps through MCP servers
- Spark schedules can now react to events and continuously monitor tasks
- Notebook can generate short form videos from source material
- AI Studio now has stronger image models, richer app building options, and direct agent creation
Individually, each update is useful. Together, they point to something bigger. Gemini is becoming more than a chat interface. It is becoming a system for building assistants, automations, apps, and media workflows in one place.
Gemini Spark now supports custom MCP apps
This is the headline update for me.
Inside Gemini Spark, there is now a Connected Apps area where you can add custom apps. The key detail is that these custom connections work through an MCP server. If you already have an MCP server for a tool or service, you can plug it directly into Spark.
That is a very big deal because it means Spark is no longer limited to a fixed set of integrations. You are not just waiting for Google to decide what gets connected next. You can begin wiring Spark into the systems that actually matter for your workflow.
In practical terms, this means Spark can become a much more useful always-on AI assistant. Instead of just answering prompts, it can potentially operate across the apps and services you care about.
What this changes
- Custom tools can be connected through MCP
- Spark can become a more capable 24/7 assistant
- You can build workflows around your own app stack instead of only Google’s defaults
- It creates a foundation for more advanced automation and monitoring
The important limitation
There is one caveat you absolutely need to know.
Custom MCP apps currently work with Gemini Spark, not regular Gemini chat.
That means if you add a custom app connection, you should not expect to immediately use it inside a normal Gemini conversation. Right now, this capability is specifically tied to Spark.
That distinction matters because Spark is being positioned as the more agent-like environment. It is where Google seems to be pushing persistent tasks, background actions, and ongoing automations. So while it may feel limiting today, it also tells you where the product is headed.
What regular Gemini chat supports right now
Regular Gemini chat still focuses on a defined group of connected services, including:
- Google Workspace
- Google Search services
- YouTube
- YouTube Music
- Google Business Profile
- GitHub
- Google Flights
- Google Hotels
- Canva
- Contacts
- Instacart
- OpenTable
- Verify AI
Google is also rolling out more integrations such as Canva, Dropbox, Instacart, OpenTable, and Zillow Rentals. So the ecosystem is expanding, but the custom MCP flexibility is the real unlock.
Gemini Spark scheduling just became much more powerful
The next update is easy to miss if you have not looked closely at Spark schedules recently.
Scheduling used to be more about recurring tasks. Now it has evolved into something much closer to a lightweight automation engine. Spark can:
- Run tasks on repeat
- Respond to events
- Continuously monitor for changes
- React based on ongoing conditions
That is a major upgrade.
A real example: monitoring home listings
One example that shows how useful this is involves real estate monitoring.
You can give Spark a request like this:
- Alert me whenever new homes are added in the Bozeman, Montana area
- Keep the price under $1 million
- Require more than two acres of land
- Keep it within 20 minutes of Main Street
From there, Spark can interpret the task, identify that it is relatively complex, and build a suitable monitoring schedule around it. It can show progress, reveal which skills it is using, and define exactly how it plans to search.
In the example, Spark set up monitoring logic around sites like Zillow and Redfin and generated the search criteria needed to periodically check for matches.
That is not just a reminder. That is ongoing task execution with clear conditions.
Why this is a bigger deal than it sounds
This update means Spark is becoming useful for a whole class of tasks that usually require separate automation tools or custom scripts.
You can use it for:
- Market monitoring
- Listing alerts
- Recurring research
- Event-based checks
- Web-connected follow-ups
And because Spark can use the web and connected apps, the automation is not boxed into a tiny internal environment. It has a much wider working surface.
Editing and controlling schedules
Another nice touch is that these schedules are manageable after setup. You can:
- Edit with Gemini
- Run a task immediately
- Pause it
- Delete it
If you need the monitoring to be more specific, the answer is simple. Tighten the criteria. Be more explicit about timing, triggers, or conditions. Spark is clearly moving toward more nuanced task handling, and this scheduling system is one of the strongest signs of that shift.
Notebook can now create short form content from your sources
This one is especially useful if you create content or repurpose long form material.
Inside Notebook, you can now generate short form videos directly from a selected source. The workflow is surprisingly straightforward:
- Open a notebook
- Select the source or sources you want to use
- Go to the video overview area
- Choose the shorts creation option
- Select a suggested topic or enter a custom topic
- Generate the short
Notebook then produces short videos based on the material, often around the one-minute mark. In the example, outputs landed around one minute and change, which is ideal for many social platforms.
What the generated short includes
The generated content is not just a trimmed excerpt. Notebook can assemble a more complete short form package with:
- On-screen text
- Visuals and generated imagery
- Structured pacing for short content
- Topic-focused framing based on the source
That dramatically reduces the friction involved in repurposing longer material into short content for platforms like YouTube, Instagram, TikTok, and Facebook.
Iteration is built in
If the first version is not quite right, you can refine it. There is an option to view the prompt and resources, and you can iterate by changing the topic or adjusting the direction.
That matters because good short form content is rarely just about summarizing. It is about focus. A single long video or article may contain several strong angles, and Notebook now helps turn each one into a separate asset much faster.
Why this matters for content workflows
Repurposing content is usually tedious. You have to identify clips, rewrite hooks, create captions, source visuals, and format everything correctly. Notebook is clearly trying to compress that process.
If you have long form videos, articles, research notes, or custom source material, this feature can help turn those into short form pieces in minutes instead of hours.
Suggested image: a screenshot of the Notebook shorts generation interface with alt text like Google Notebook short form video generation from source material.
AI Studio adds Nano Banana 2 Lite and expands image generation options
The rest of the major updates are inside Google AI Studio, and there is a lot going on there.
One standout change is a new image generation and editing model called Nano Banana 2 Lite.
On paper, it may sound like a smaller version of an existing model. In reality, that can be exactly what makes it valuable.
Why Nano Banana 2 Lite is interesting
Nano Banana 2 Lite is positioned as the smallest and most cost-effective image generation and editing model in the lineup. The important comparison is that it is not dramatically behind Nano Banana 2 in quality, but it is meaningfully faster and cheaper.
The model supports:
- Image and text input
- Image-only workflows
- Image generation
- Image editing
The advantages called out include:
- Performance that stays relatively close to the larger model
- Latency that is much lower
- Pricing that is roughly half the cost
If you are building with the API, running large numbers of generations, or trying to optimize credit usage, this is not a minor improvement. It is a practical upgrade that can change what is financially viable.
Who should care about this model
- Developers using image APIs at scale
- Teams that need faster iteration
- Builders balancing cost and quality
- Anyone experimenting with image workflows inside AI Studio
Suggested infographic: a comparison chart for Nano Banana 2 Lite versus Nano Banana 2 with alt text like Nano Banana 2 Lite pricing latency and image quality comparison.
AI Studio app building is getting seriously capable
AI Studio is no longer just a playground for prompts and models. It is becoming a full builder environment.
You can now create a wide range of projects, including:
- Android apps
- Web apps and websites
- Chrome extensions
- Mobile app experiences
And the list of capabilities you can plug into those projects is honestly kind of wild.
Google service integrations available in app building
- Google Drive
- Google Sheets
- Gmail
- Google Calendar
- Google Docs
- Google Slides
- Google Tasks
- Google Chat
- Google Forms
- Google Keep
- Google Meet
- Contacts
Additional AI capabilities you can build in
- Text to speech
- Music generation
- Database and authentication
- Create and edit images
- Voice conversations
- Animate images into videos
- Use Google Search data
That combination of app scaffolding plus AI functionality makes AI Studio far more practical for real product experiments. You are not just testing prompts anymore. You are assembling applications with native access to useful services.
Design variations make rapid prototyping easier
Another update inside AI Studio makes iteration much easier for anyone building a site or app interface.
When editing a project, you can now use tools like:
- Annotate
- Focus
- Design variations
The design variations feature is especially useful because it can automatically generate multiple visual directions for the same app or page. In the example, a pizza site was turned into different design treatments with different color schemes and layouts.
This is powerful for a few reasons:
- You can compare styles quickly
- You can describe a design direction and get options back
- You can iterate visually without leaving the builder
- It shortens the distance between idea and prototype
For solo builders and small teams, that is a huge productivity boost.
AI Studio now lets you build agents directly
One more major addition inside AI Studio is direct agent creation.
In the model selection area, there is now an agents option that lets you build agents inside the platform itself. Templates and examples include things like:
- Document processor
- AI talk radio
- Customer support
- Repo maintainer
Or, if you want something custom, you can simply describe the kind of agent you want to build.
This runs through the agent preview experience and points toward a future where AI Studio is not just for testing models but for building deployable AI workers and assistants.
Why this matters
People often jump between separate tools for prompting, prototyping, integration, interface building, and agent setup. Google is gradually compressing those layers into one place.
That means AI Studio is becoming useful for:
- Rapid proof of concept work
- Internal tool creation
- Agent experimentation
- Building AI-powered user experiences
- Connecting Gemini capabilities to practical applications
The big picture: Gemini is becoming an automation and creation platform
If you zoom out, the pattern is pretty clear.
Gemini Spark is moving toward persistent automation with custom app connectivity and reactive scheduling. Notebook is becoming a content repurposing machine. AI Studio is turning into a genuine development environment for apps, images, interfaces, and agents.
That matters because the value of AI is no longer just in asking clever questions. The value increasingly comes from building systems that keep working after you stop typing.
And that is exactly where these updates are pointing.
FAQ
Can custom MCP apps be used in regular Gemini chat?
No. At the moment, custom MCP app connections are available for Gemini Spark, not regular Gemini chat. Standard chat still supports a more limited set of built-in integrations.
What is the main benefit of custom MCP support in Gemini Spark?
It allows Spark to connect to custom apps and services through MCP servers, making it far more flexible as an always-on AI assistant and automation layer.
What can Spark schedules do now that they could not do before?
Schedules can now do more than repeat tasks. They can respond to events, continuously monitor conditions, and react based on changes over time.
Can Notebook generate short videos from long form content?
Yes. Notebook can take selected source material and generate short form videos with text, visuals, and topic-based structure, making content repurposing much faster.
Why is Nano Banana 2 Lite important in AI Studio?
Because it offers a strong balance of quality, speed, and price. It stays relatively close to the larger model in performance while being faster and more cost-effective for API and credit-based workflows.
What kinds of things can be built inside AI Studio now?
AI Studio can now be used to build apps, websites, Android experiences, Chrome extensions, image workflows, voice features, and even custom agents.
Final thoughts
If you have not checked Gemini lately, now is a good time. These updates are not just cosmetic. They change how useful the platform can be for real work.
The custom MCP support in Spark is the one to keep the closest eye on. That is the feature with the biggest long-term upside. But the scheduling improvements, Notebook shorts, and AI Studio upgrades all make the ecosystem much more capable right now.
If you are building AI workflows, experimenting with agents, or trying to create content faster, there is a lot here worth testing.
Share this with someone who is sleeping on Gemini, and if you are exploring these features already, compare notes and keep pushing on the edges. That is usually where the best use cases show up first.
