|
October 2026 |
|
|
Lyzr Studio 2.0 · What is new
Superflow learns to listen.
This month you can build a Superflow just by describing it. Your agents can carry more context, and you get clearer control over which model answers and what it costs. Here is everything that shipped.
|
|
|
01 · Superflow · New
Describe the flow. Superflow draws it.
Building a Superflow is now a conversation. Open Build with AI, say what you want in plain language, and the flow appears on the canvas node by node with a short summary of what it does. Run sample input to watch it work end to end, then keep chatting to change it. Any step can be undone with Ctrl+Z.
|
One prompt, seven nodes, a prospect analyzer ready to test
|
|
|
|
|
Also in Superflow · New
More it can do. More it can remember.
|
|
A new look
Cleaner nodes, a node library you can search or browse by category, and a calmer canvas to build on.
|
|
Versioning
Keep track of how a flow changes over time, so you can keep improving it with confidence.
|
|
Knowledge Base, Loop Back and Loop Until nodes
Pull answers from a knowledge base mid flow, send work back to an earlier step, or repeat a step until a condition is met. Together they make complex orchestration possible on a single canvas.
|
|
A memory module
Add memory and a flow keeps track of what it has already seen, then uses it to decide what to do next.
|
|
Create agents inside Superflow
Create an agent, set it up and wire it into the flow without leaving the canvas.
|
|
|
02 · Agents · New
More context. Fewer detours.
|
|
Multiple global contexts per agent
An agent is no longer limited to one global context. Keep brand voice, product facts and policies as separate contexts, and attach as many as the agent needs.
|
|
Global contexts and skills from the builder
Create both on one page, right inside the agent builder, without switching screens or losing your place.
|
|
|
03 · Models and cost · New
Compare models side by side.
In Playground, send one message to several models at once and read the answers next to each other. Token count, total time and time to first token sit under every response, with the fewest tokens and the fastest reply flagged for you. When one stands out, choose Use this model.
|
|
|
04 · Knowledge and teams · New
Sync what matters. Share what works.
|
|
Folder level SharePoint sync
Sync a SharePoint knowledge base folder by folder, so you control exactly what gets synced.
|
|
Share skills with your team
Share a skill with one or more teammates. Build it once, and everyone who needs it can put it to work.
|
|
In case you missed it · Live today
Shipped earlier, worth a look.
|
|
Auto model routing
Turn it on and every request is classified as simple, medium, complex or reasoning, then served by the model you assigned to that tier. Greetings stay quick and cheap, deep analysis gets the depth it needs, and each tier shows its cost and speed so the trade off is clear.
|
|
|
|
FinOps rate limiting
Control cost at every level of your organization. Set one ceiling for everything, a budget per team or client, limits for a single agent, or limits for each person.
|
|
Jev decision model in Superflow
Jev, the classifier only model from TypeSafe, is available in Superflow nodes. Reach for it wherever a step needs to classify rather than generate.
|
|
|
Everything here is live in Lyzr Studio today. If you build something with it, or something gets in your way, just reply to this email. We would love to hear from you.
The Lyzr Studio team
Open Lyzr Studio
|