August 2026 monday.com Product Updates
monday.com's August 2026 updates: another release that is right on the money
Native Python in workflows, a proper database behind Vibe apps, as well as mobile deployment.
There's a wall most work platforms run into. You can configure them a long way, and then you hit something the configuration doesn't cover and you go outside the platform to solve it. Middleware, a script somewhere, a separate database, a spreadsheet that nobody admits exists. Every implementation I've worked on has some sort of version or interpretation of that.
August is the month monday took a serious run at that hurdle. You can write Python inside a workflow. Vibe apps can sit on a dedicated database holding a million records. You can deploy a Vibe app to someone's phone as their home screen. Vibe apps work with multi-level board hierarchies. And when someone deletes something important, you can get it back yourself.
There are nineteen updates in total and plenty of them are about agents, which I'll come to. However the ones that have my attention are those that widen what you can build without leaving the platform.
Here's the breakdown and as always this is roughly in the order I'd rank it.
The Workflows Code block
This is the most important aspect of the release and I don't think its description does it justice.
You can now add a Code block at any step inside a monday workflow and write custom Python in it, without leaving the builder. Errors surface in the run history in real time, so you can debug a failing step from inside the workflow rather than guessing. The block can also be created and wired up through monday's MCP.
Now, every non-trivial build eventually needs logic that doesn't fit into standard blocks. A data transformation with real rules behind it. Formatting that has to be exact because something downstream depends on it. Conditional logic with more than a couple of branches. Until now, that meant going outside monday, usually into middleware, and then maintaining a piece of your process in a second system. Two places to check when something breaks, two places to update when the requirement changes, and a dependency that somebody has to keep paying for and remembering about.
Having it as a native step collapses that back into one place.
It matters most for migrations. When we move a client onto monday from another platform, the configuration work is usually the easy part. The thing that stalls a migration is the handful of processes with real logic in them, the ones somebody built years ago in a system that let them write code. Those are what force either an ugly compromise or a middleware layer that undermines the case for consolidating in the first place. A native code step covers most of them. monday's own view is that this is the most useful capability for accounts migrating in... and I agree with them greatly.
The MCP detail is worth noting too. You can build and wire the block from an external AI tool rather than in the builder.
Paired with this: triggering workflows from any external system. Workflow builders can now fire a monday workflow from outside using a secure inbound webhook, whether that's a CRM, a ticketing tool, a survey platform or something custom. Task creation from Stripe, Typeform or Shopify becomes something you set up in monday rather than somewhere in between.
Take these two together and they change the calculation on middleware. I said last month that I wanted to watch how the MCP block affected where we reach for tools like Make, and that I wasn't ready to call it. Inbound webhooks plus native Python is a much more serious answer than the MCP block was on its own. Data comes in from outside, logic runs natively, work happens in monday. I'm still not calling middleware dead, because there's a lot it does well and plenty of integrations where it remains the right choice. The range of builds now however that need it has narrowed noticeably in two months.
Vibe: it now has a real database underneath it
Vibe is the area I'm most consistently excited about and this month it got what it's been needing the most.
Vibe DB. Builders can provision a dedicated database inside a Vibe app in a few clicks. It holds up to a million records and monday quotes query speeds around 30 times faster than a monday board. Data persists across app rebuilds and you can migrate existing board data into it.
Treat those figures as monday's own and worth testing rather than assuming, but the direction is unmistakable. Boards are a good general purpose structure and a poor high-volume database. That has been the real ceiling on Vibe. Any app involving a lot of records, or needing quick queries, or holding data that had to survive the app changing, wasn't viable. So Vibe apps stayed small and single-purpose, which is fine, but it isn't what the tool is capable of.
With a proper database behind it, the category now changes. Custom CRMs. Audit and event logs. Session or activity tracking. Operational dashboards over meaningful volumes of data. Internal tools that a business would otherwise have gone out and bought or commissioned.
The persistence across rebuilds is the one I'd underline, because it's the least obvious and the most practical. If iterating on your app risks your data, you stop iterating, and an app you can't safely change is an app that slowly stops matching the process. Reliable persistence is what makes a Vibe app something you maintain rather than something you built once and not really touch again.
Vibe on mobile. Admins can set a mobile home screen for specific users or teams, pointing them straight at a chosen Vibe app when they open the monday mobile app. It's configured through the Mobile Experiences setup under Admin, where you pick the app, assign it to users and name it. Those users then land directly in their app with no navigation.
This is a bigger deal than it sounds for anyone with people away from a desk. Field engineers, warehouse and stockroom staff, drivers, site teams. Mobile adoption fails on navigation more than anything else in my experience. If the thing someone needs is two or three taps deep inside a general purpose app, they won't use it and you end up with a process that works on paper and not in practice. Opening the app and being in the tool built for your job removes that entirely.
Combine it with the barcode and QR scanning that arrived in Vibe a couple of months back and there's now a genuinely credible path for operational, on-site work: scan on a phone, into a purpose-built app, backed by a database that can handle the volume. This is an enormous step in the right direction.
Multi-level subitem boards. Vibe apps now support boards using nested subitems, so phases to tasks to sub-tasks, with the app able to read, create and update items at every level of the hierarchy from one place. Multi-level hierarchies are standard in any larger board design, so this removes a common reason a Vibe app wasn't an option for a serious project structure.
Plan Mode.Rather than starting from a blank canvas, you describe what you want in the Vibe chat and the AI asks clarifying questions, then produces a structured plan document covering the app's goals, structure and key capabilities before anything gets built.
Anyone who has built software will know why this helps. The expensive mistake is never a bug, it's building the wrong thing confidently and finding out late. A structured conversation up front costs a few minutes and prevents a lot of rework. It's also the difference between someone getting a usable app on their first attempt and giving up after three goes.
Build Vibe apps from any AI tool. You can create, edit, publish and manage Vibe apps from any MCP-compatible AI client, including ChatGPT and Claude, without opening monday at all.
Password protection for public apps, so you can share a live interactive app with people outside your account without leaving it open to anyone who has the link. Straightforward and necessary if you're going to share anything real externally.
Restoring deleted portfolios
Admins and account owners can now restore a deleted portfolio themselves within 30 days of deletion, without going through support. Portfolio projects that reconnect successfully are restored quietly and only the ones that can't reconnect get surfaced, so you aren't wading through noise to find the actual problem.
I'm pleased to see this for a specific and slightly painful reason. We've had clients delete portfolios, and rebuilding one is a very time consuming job. A portfolio is a set of connections to projects, and reconstructing those connections by hand while working out which projects were in there and how they were configured is slow, careful work that produces nothing new. It's the sort of task that can eat up a whole day and leaves everyone feeling worse off.
Being able to undo it yourself, immediately, without raising a ticket and waiting is worth more than the feature description suggests. Most of the cost of that kind of mistake isn't the restoration itself, it's the delay while you wait for someone else to fix it and the reporting that's broken in the meantime.
The wider point is that self-service recovery changes how people behave. Teams that know a mistake is reversible work faster and restructure things when they should. Teams that know a mistake means a support ticket and a week of pain leave things alone, and accounts can slowly fossilise.
Agents: more capable & more governable
Agents got a substantial amount of attention this month too.
They can do more on their own. You can duplicate any agent you own or have access to, which matters once you have a few since most agents in an account are variations on each other and rebuilding each from scratch produces inconsistent results. Agents can now edit columns with restricted permissions when explicitly authorised, either by making the agent a board owner or granting access in the column settings, which removes a manual step from any process touching sensitive fields. They connect to Dropbox for reading, creating and sharing files, and to Canva design assets through MCP. Also, image generation is now available to everyone.
Setting them up got much simpler. You can configure tools, set guardrails and manage sensitive permissions from inside the agent's chat. It offers the right tool for the task, treats read access differently from write and delete, and asks for explicit confirmation before activating anything sensitive. Tool configuration has been the point where non-technical builders give up, so doing it conversationally is what lets people outside IT now build their own agents.
Bring Your Own Agent also got simpler. You can connect external agents built on natively supported platforms such as Claude or Copilot Studio, or bring in a custom-built agent through a webhook. Invited agents become governed entities with their own identity, separate from the person who invited them, and can be chatted with, mentioned, assigned tasks, triggered and monitored like anyone else. Whilst simultaneously their actual instructions and skills stay on the external platform. The webhook route is the useful addition, because it means you aren't restricted to the platforms monday has built native support for.
And the controls arrived alongside. Admins get layered permissions for MCP: an account-level master switch with per-connector sub-permissions, workspace level scoping, and on Enterprise a role level layer defining which agent types each role can access. They can monitor tool usage across every agent in the account and revoke a tool instantly. There's a connection graph showing which users are running which agents and what those agents are connected to. As well as this, there's an account-wide toggle controlling whether people can message their monday agents from WhatsApp or Telegram.
Most conversations about scaling agents stall on governance rather than capability, and a single on-off switch forces an all-or-nothing decision that organisations usually resolve as "off". Being able to say MCP is on, but only these connectors, only in these workspaces, and only for these roles, turns a blanket refusal into a very useful tool.
Resource planning keeps getting better
Resource and project managers can now configure individual work schedules, meaning working days and hours per week, for each person. On top of that there are regional holiday calendars, and personal time off such as annual leave and sick leave, all set at the individual level. Those three availability layers feed through in real time to the Workload widget, the Resource Planner and the Capacity Manager.
Resource management is one of monday's most underused capabilities, and I've thought that for a while. Plenty of accounts have it available and aren't touching it. So watching monday put real development effort into it two months running is one of the more encouraging trends on the platform. Last month brought portfolio level utilisation, planned against actual across every project. This month fixes the foundation underneath it.
The reason it needed fixing is that capacity planning is only as good as its assumptions about who's available. A planner that treats everyone as full-time on a standard week is wrong about anyone part-time, wrong about anyone in another country with different public holidays and wrong about anyone on leave. In most teams that's a significant share of people and once a plan is visibly wrong about availability, people stop trusting it and go back to a spreadsheet.
Three independent layers, per person, reflected live in the planning views, is what makes the numbers defensible. If you've looked at monday's resource management before and decided it was too blunt for how your team actually works, this is a fair point to look again.
Also shipped
Workspace 2.0. New workspaces are now built as places where people and agents work together. Each one comes with a built-in workspace agent that reads boards, answers questions and takes action, plus shared knowledge you can load with files and connected boards, a group chat where the team and the agent talk in one place, and a folder where everything the agent produces gets saved. The important caveat however, and it's the first thing to tell clients: these parts apply to new workspaces only, not existing ones. If your account is already set up, they won't be retrofitted. This is worth a deliberate decision rather than letting it drift, since it affects whether a new function gets a fresh workspace or goes into an existing one.
Channel health visibility in monday Service. Admins get real-time status on every intake channel, alerts when one breaks, and the ability to transfer ownership themselves. A silently broken support channel is one of the worst failure modes in a service operation, because nothing looks wrong until a customer asks why nobody replied.
Manual workflow triggers have reached full release. There's a Trigger Now button inside the workflow, so you can start one on demand with no status column or board-based trigger. This began rolling out last month and is now fully available, so if you looked for it and couldn't find it, it is now there
What I would do about this month
Look at the Code block if you have a process that didn't fit.If you've got logic sitting in middleware purely because monday couldn't do it, or a requirement you had to compromise on, that constraint has changed.
Reconsider the Vibe apps you decided weren't viable.If an idea was ruled out on data volume, query speed, mobile access or board hierarchy, all four of those objections moved this month.
Fix your availability data.Per-person schedules, holiday calendars and time off are the foundation the rest of capacity planning sits on. Setting them up is unglamorous work and it's the difference between a resource plan people use and one they ignore.
Decide your Workspace 2.0 approach.Since the new features only apply to new workspaces, this needs a decision now rather than later.
And as ever, don't try to adopt nineteen updates at once. Pick the one or two that match a problem you are already facing.
How The SaaS JEDI helps
The SaaS JEDI is a monday.com implementation partner working with clients across the UK and South Africa. On this release we can help you:
Build workflow logic using the Code block, and consolidate processes currently sitting in middleware
Set up inbound webhook triggers from your CRM, billing, intake or e-commerce systems
Build Vibe apps on Vibe DB for higher-volume use cases, and deploy them to mobile for field and on-site teams
Design Vibe apps against multi-level board hierarchies for larger project structures
Set up resource management properly, including per-person schedules, holiday calendars and time off, and portfolio-level utilisation reporting
Design, build and govern monday agents, including tool guardrails and authorised access to restricted columns
Configure agent and MCP governance: layered permissions, tool controls, connection visibility and messaging channel settings
Plan a Workspace 2.0 approach that works with your existing account structure
Train your teams so this gets used rather than just announced and then forgotten
If you'd like to work out which of August's updates are worth prioritising for your account, get in touch.
Webinar: NEW AI in CRM with Agents
We're running a session on how agents change the way a CRM gets used day to day, what's worth automating, where a human needs to stay in the loop as well as what we've learned building agents for clients:
September 30th, 13:00 - 14:00 BST
Sign up here using this link:
If your CRM data is only as good as the last person who remembered to update it, you need to come to this one.
Stephen MacLennan
Founder, The SaaS JEDI
