Make Waves '26: What you can use today and what is still only on the waitlist
At Waves '26 on October 20, Make announced a long list of features for AI agents, but only some of them work today: the Maia upgrade, Knowledge, Agent orchestration, Custom Roles, Private Spaces, Make Grid, Event subscriptions and Audit logs. Human in the loop and MCP toolboxes are announced, Process Intelligence and MakeOS are waitlist only. In Germany, task mining needs the works council on board.
The Waves '26 announcement is mostly a roadmap. Make groups it into four areas (Discover, Build, Govern, Observe) plus a new product called MakeOS. According to the Make blog post from October 20, many items are marked "Coming soon", with no date and no price. So the question that matters for your planning is simple: what can you switch on today?
What is available in Make right now?
Eight features are usable today according to Make. The rest are announced or only open via a waitlist. The table separates the two:
| Feature | What it does | Status per Make |
|---|---|---|
| Maia upgrade | AI assistant plans first, remembers preferences, reads images, suggests error fixes | available |
| Agent orchestration | one AI agent hands subtasks to subagents | available |
| Knowledge | shared company context for people and agents | available |
| Custom Roles | permissions assigned one by one, agents inherit their owner's permissions | available |
| Private Spaces | experiments and credentials stay private, even from admins | available |
| Make Grid | map of all scenarios and agents, including other platforms, with full text search | available |
| Event subscriptions | alerts on events, such as agent runs that finish but go wrong | available |
| Audit logs for agents and MCP | immutable record of who changed what | available |
| Human in the loop | scenario pauses and waits for approval via Slack, Teams or app | announced |
| Run up to a module, Data pinning | test from the middle of a scenario with saved test data | announced |
| Functions, Endpoints | agents run code or may only use single API actions | announced |
| MCP toolboxes | fixed toolset per agent for Claude, ChatGPT and others | announced |
| Organization policy, Locked connections | set allowed apps and AI models centrally, lock connections | announced |
| Agent evals and AI spend | measure whether agents improve and what they cost | announced |
| OpenTelemetry | stream Make data into your existing monitoring tools | later |
| Process Intelligence, MakeOS | process analysis and a dedicated workspace for AI agents | waitlist |
MCP stands for Model Context Protocol, the open standard that lets AI assistants like Claude or ChatGPT call outside tools. For what that looks like with your own app, see our post on the Supabase MCP server.
What does the Maia upgrade change day to day?
Maia, the AI assistant in the scenario editor, can now wait for your approval before it builds anything. In the new Plan mode, Maia asks all its questions up front and only starts building once you confirm the plan, as Make explains. The existing Build mode stays the default. On top of that:
- Memory: up to 50 entries of up to 200 characters each, per user and team, invisible to colleagues
- Images: up to three per session (PNG, JPEG, WebP, 1.5 MB each), for example a whiteboard photo of your process; Make says it does not store them
- Resolve error: Maia reads the data from a failed run and suggests a fix that you review before saving
Our take: Plan mode and Resolve error are the two features that save time right away. Still, an AI fix is only a suggestion. In scenarios that create invoices or send emails, test every change against real test data before you save it. Make's post mentions neither plans nor credit usage for the new features, so check your account before the whole team starts relying on Maia.
What is Make Process Intelligence and when is it coming?
Process Intelligence is meant to show which work is worth automating, but for now it is waitlist only. According to Make, it is built on Celonis technology; Make itself belongs to Celonis, as the footer of the Make website shows. It has two parts:
- Process mining analyzes event data from your systems of record and shows where cases get stuck. You can ask questions in plain language, such as why a case is delayed.
- Task mining records how people work on the desktop and in the browser (via a Chrome extension), groups those actions into tasks and estimates how often each one is done by hand per month. "Build with Maia" then turns a task into a scenario.
Make gives no pricing, launch date or regions.
What does task mining mean for companies in Germany?
Task mining logs screen work, and Germany has firm rules for exactly that. Under Section 87(1) no. 6 of the Works Constitution Act (Betriebsverfassungsgesetz, or BetrVG, the German law on employee co-determination), the works council (Betriebsrat, the elected employee body) has a say in technical systems designed to monitor employee behavior or performance. Our take, not legal advice: a tool that analyzes screen work by time spent will hardly run in a company with a works council without a works agreement. Data protection questions come on top.
Make does ship some controls for this. You choose which apps are in scope, and according to Make nothing outside them is collected. Capture depth ranges from "app was used" up to window titles, text input and screenshots can be turned off, usernames are hashed, and employees can see that capture is running and pause it. That is a solid starting point for the conversation, but it does not replace it.
What does this mean for you?
This week, the best move is to clean up with what is already there. Three steps:
- Set up Event subscriptions: get alerts for the events you care about the moment they happen, including agent runs that finish but go wrong.
- Review permissions: use Custom Roles to assign permissions one by one. Agents inherit their owner's permissions, so never build agents under an admin account.
- Open Make Grid: take one look at which scenarios nobody remembers anymore, and deactivate old ones instead of deleting them.
If you want to test task mining, join the waitlist and talk to your works council and data protection officer in parallel. Torn between Make and n8n? We ran the cost numbers in Make vs n8n. We would only plan around the announced features once they show up in your own account. If you need help cleaning up existing scenarios, get in touch.
Frequently asked questions
Does the Maia upgrade cost extra?
Make's announcement lists no plans and no credit usage for Plan mode, Memory or Resolve error. Check your account or ask Make before you roll the features out to the whole team.
What is MakeOS?
MakeOS is a new Make product: a workspace where AI agents handle tasks across the company. They work within rules that IT approves once in advance; if an agent tries to go past its limits, MakeOS blocks it, asks a person and logs the attempt. So far there is only a waitlist and no launch date.
Can I use Human in the loop today?
No. According to Make, the approval pause via Slack, Teams or app is announced but not yet available. Until then, an extra step in the scenario with an approval email or a form does the job.
Does the works council point also apply to process mining?
Process mining analyzes data from business systems. As soon as that data allows conclusions about individual employees, similar questions come up. Clarify it case by case with your works council and data protection officer.
Sources
Tell us what you're planning. You'll get an honest assessment within 24 hours.
More articles