As folks know by now, we have been releasing feature updates to our products every few weeks in the form of “Waves.” However, as we continue to accelerate our development, we have come to the point where covering everything we are shipping in a single day or a single post just would just be too much.

So, for Wave 8, we are doing something slightly different: instead of releasing everything on a single day, we are going to be shipping the new capabilities across this entire week, each day focused on a different theme. Get ready, since this is just day one!

Today’s theme: Teams & Enterprise.

The quick tl;dr on what we’re launching today:

Windsurf Reviews: GitHub app that can review PRs and edit titles/descriptions given code review guidelines.
Knowledge: A refresh of our context awareness engine that can now also read from Google Docs.
Conversation Sharing: A simple way for team members to internally share Cascade conversations that worked for them
Teams Deploys: Securely deploy apps directly to an internal, admin controlled Netlify account
Updated Admin Analytics: A full refresh of our admin analytics portal for building a business case around Windsurf adoption

Why Teams & Enterprise

We have always been a company that values and prioritizes enterprise use cases. Why? For one, developers at enterprises operate on much larger scale codebases and much more intricate tasks than the hobbyist developer, and so they reflect some of the hardest challenges we could tackle as an AI platform for software engineering. And second, there are over 30 million professional developers globally at enterprises that drive the vast majority of global economic value and societal impact, and this number is only growing. We add the most value by adding the most value to these folks. The problems are challenging and the opportunities are large.

That being said, the folks rolling out these tools within teams and enterprises care about things more than just the value that an individual developer gains while writing code, and often have considerations beyond just the scale and complexity of the codebases. These include (but are not limited to):

How do we hook into data sources beyond just code and add value to parts of the overall software development lifecycle beyond what is done in the IDE?
How do we maintain organizational best practices and match existing standards around style and syntax?
How do we respect organizational security and compliance standards?
How do we enable effective adoption of the technology across the organization, via analytics and dissemination of best practices in using the tool?

While we have always thought about and worked on these problems, we believe the core multi-step reasoning engine that we’ve introduced with Cascade has massively expanded the frontier of possibility. As a simple example, what good would it have been to connect more data sources if we only had a single call to an LLM with a limited context length anyways? We wouldn’t have been able to add the information exposed by these connections! But now with multi-step reasoning, this becomes more interesting.

In Wave 8, we are releasing a whole slew of features and capabilities across individual developers and admins of organizations to tackle these particular kinds of questions, making Windsurf the single most advanced agentic platform for enterprise software engineering teams.

Windsurf Reviews

Any professional developer knows that part of the time is not spent doing tasks in the IDE. One of the other important parts of the software development life cycle is code review, the process where a peer checks a proposed set of code changes, leaves comments, and approves the changes before they are merged into the main “golden” branch of the codebase. This is a fundamental part of working in large codebases built by teams (as opposed to individual developers) in order to maintain code quality, best practices, and a number of other factors independent of pure “code correctness.” However, any developer will tell you that it takes time (a) as an author to come up with a clean description of the code and (b) as a reviewer to pore through the proposed changes and leave every comment, especially nits.

Windsurf Reviews is a simple concept - help both author and reviewer where the work is tedious and time-consuming, but still have a human in the loop. Windsurf Reviews are focused on the smaller issues as opposed to larger or more nuanced problems.

How does it work? Well, an admin heads over to the corresponding Windsurf Settings page here, and connects their GitHub instance, using a GitHub bot:

After an admin enables the PR Reviews and PR Description Edits toggles, the full power of Windsurf Reviews is ready for all users in the organization, whether or not they have a Windsurf seat!

An author can specify /windsurf at the start of the description or just as the title when creating the PR, and Reviews will automatically generate a description based on the diff and context:

On the reviewer side, Reviews will be automatically triggered when the PR is marked ready for review, and can also be triggered by leaving a comment of /windsurf-review:

Both of these features can also follow centrally defined rules (also from the admin page), so that Windsurf Reviews can follow centralized style guides and best practices for code review:

Add additional context for Reviews to follow

Note that Reviews does require data retention on Windsurf’s end and must be opted-in by connecting the GitHub bot. This includes basic telemetry for review comments such as timestamp, repository name, GitHub username of the PR author, comment ID, reaction count (thumbs up/down), and the status of suggested commits.

We are launching Reviews in beta for Teams and Enterprises. What does beta mean in this context?

We are currently offering Reviews for free so that we understand usage patterns. Reviews will eventually adopt some sort of pricing structure, but this will only come once we clearly understand the costs and value.
Because we are offering Reviews for free, there are certain global org rate limits on number of reviews per day (500 reviews / mo), and a limit on the max size of diff that we will process (~350k total characters across max 30 files).
Reviews only currently work for organizations that have their source code in a GitHub Cloud instance. Other source code management setups to come after initial value is proven out.
Because of the Windsurf-side data retention requirement, Reviews will not be available to Hybrid customers while it is in beta. Once we get positive feedback on Reviews, we will bring Reviews to Hybrid with the data retention happening in the customer’s tenant.

We are very excited about this version of AI in the code review process because it keeps a human in the loop to trigger the AI, and isn’t set out to overstep or tackle very complex logic or nuances. The goal is to assist developers by removing the parts of the code review process that are tedious.

Knowledge Base

Ok, back to the Editor. One of our core differentiations from other AI tools for the last couple of years has been our ability to deeply understand existing codebases at enterprise scale, thereby being able to ground our AI responses. However, any enterprise developer knows that codebases do not contain all of the relevant context for their work - a lot of relevant information lives in docs, tickets, and more.

When these AI features were restricted to single LLM calls, connecting these additional data sources simply was not that interesting. Even if we connected more data sources, we still had a limited context length to put this information into, and so it was unclear how much of the now-connected data could be incorporated into the responses. However, with the new paradigm around multi-step retrieval and reasoning, connecting more data sources now has a lot more potential.

We are expanding the idea of “context awareness” of a codebase to an idea of a “knowledge base” of multiple sources. The first new source will be documentation, and for this beta launch of Knowledge, we are integrating with Google Docs.

Similar to Windsurf Reviews, this starts with the admin heading over to the corresponding Windsurf Settings page here, and following the instructions to connect their account to Google Drive, and then adding items:

Note that this is using the admin’s Google account, so all requests made to the Google Drive API will be using the admin’s API key. For the beta launch, there are no additional access controls, so all docs manually added as items can be referenced by any Windsurf user, whether or not they actually have access to the original doc (as this works as long as the admin has access to the doc). Therefore, admins should only add docs that they want everyone to be able to reference, and to have additional peace-of-mind of not giving access to restricted files to the broader user base, companies may consider creating a separate “service account” Google account that they explicitly add to the selected subset of docs, knowing that all Windsurf users will have access to these. We use the Google Drive API drive.file scope to limit access to only the files selected to make sure that Windsurf does not have access to unselected files.

Once added, these docs will be used by Cascade, either automatically:

or by explicit at-mentions using @knowledge:

In order to assist Cascade in identifying whether a doc would be relevant for a query, an admin can add additional textual descriptions of when the file should be retrieved:

Note that this is available today as a beta feature to all Teams and Enterprises, excluding Hybrid customers. There is currently a caching data retention component, where we store the title and contents of the files. Thus, to respect the data retention agreements for Hybrid, we currently cannot provide this feature. Once we build confidence in this approach to Knowledge, we will rapidly bring the data retention layer to Hybrid customers’ tenants. As part of this beta launch, there is also a maximum of 50 documents allowed to be linked.

Now, you may ask how this differs from just having an MCP server that exposes Google Docs, which is fair. We currently see a few axes:

We expose explicit control and curation of the files
There is the ability to add guidance when to use the files in a way that Cascade is familiar with
There is chunking of unstructured text that is more aligned with how Cascade currently processes unstructured text (as with parsing webpages as part of web search)

We still believe strongly in standard protocols like MCP as we don’t have a desire to build every data connector in house, but we find value in doing so for some of the data sources that we (a) view critical to the general software development life cycle and (b) believe we can create a differentiatedly better direct interface with Cascade.

This is the start of our foray into connecting additional data sources, and as we prove out our confidence in our approach with handling unstructured data, we will add more sources, more access control features, and more. We are excited about developing Knowledge Bases further.

One of the most pressing challenges for organizations with regards to AI code assistants is actually adoption. There is always a subset of early adopters that realize the potential of these tools, and are constantly tinkering with whatever is new (and we have a lot of new pieces all of the time), but the vast majority of developers are either slow to adopt or slow to reach near-maximal value, especially in larger organizations.

One of the best ways to “solve” adoption? Have those early adopters share with everyone else how they are using the tool for the particular use cases of their company. We knew that taking screenshots of the Cascade panel was less than ideal, so we built out native conversation sharing.

All a user has to do is click “Share Conversation” in the top right menu of a Cascade conversation, and we will provide a team-gated link that will render the entire conversation in a webpage. Only other logged-in members of that same Team or Enterprise will be able to view the contents of that link for security purposes:

Note that this beta feature naturally does require the contents of the conversation to be stored on Windsurf’s servers, so it must be manually enabled by a Team/Enterprise admin on the corresponding settings page. For the same data retention reason as with Windsurf Reviews and Knowledge Base, we are not releasing this to Hybrid customers today, but will be coming soon.

Admins also have the ability to opt-in to allowing conversations to be shared with the Windsurf team, which would assist in debugging and providing feedback on how to even better use the platform.

Teams Deploys

Back in Wave 6, we released Deploys, a one-click way to take the application you are locally iterating on and make it viewable on the public internet. This process does require the code to be uploaded to Windsurf’s servers so that we can deploy the application within our Netlify account, after which the user could claim the deploy into their own Netlify account if desired.

In this Wave, we are making it possible for a Team or Enterprise admin to specify an existing Netlify account for that organization, after which any Deploy for any corresponding Windsurf user can be directly sent to the organization’s Netlify account, by-passing the Windsurf-deployed-and-then-claimed flow for individual users:

Internally at Windsurf, we have built hundreds of internal apps using Windsurf and deploying with Deploys. Some of these have saved us hundreds of thousands of dollars in B2B SaaS spend (see video). We see Deploys being especially useful for organizations, not just for internal apps, but also for dev-adjacent roles, such as allowing PMs to share rapidly built prototypes instead of writing long product spec docs for the engineering team.

Updated Admin Analytics

This one is pretty simple. We’ve always prided ourselves on having some of the deepest, most robust, and most trustable analytics because we understand that, even though this is a technology that everyone is saying adds a lot of value, our customers still need to prove the ROI internally, whether with their finance team, executives, or board. We’ve completely revamped our analytics dashboard, provided additional views and slices, and added metrics around Cascade usage:

Admins should check out the new analytics page here.

Self-serve Teams access controls and Self-serve Enterprise tier

When we announced our new pricing system, we massively simplified all of our pricing and tiers, removing the concept of flow action credits, consolidating tiers, and making plans cheaper across the board.

As part of that announcement, we also said that a couple of things were coming soon:

The ability for self-serve Teams customers to pay an add-on of $10/u/mo for the whole suite of advanced access controls features, including SSO, rule-based access controls, and subteam analytics.
The ability to self-serve the Enterprise tier, which comes with all of the access control features, APIs to the analytics, and more credits at a cheaper cost-per-credit rate.

Well, as part of this Wave 8, these are now available! Now Teams don’t have to pay Enterprise prices if all they want is access control features, and Enterprise customers don’t have to chat with our team if they just want to get started.

That’s a wrap on Day 1 of Wave 8! Hopefully, the core goal is clear of all of these new features and capabilities. If we can provide value in the most complex software engineering environments, tasks, and requirements, then we will be able to provide value to everyone who is trying to build anything with software.

See you tomorrow, where we will release more features that everyone will benefit from, from individual builders to enterprise developers. Wave 8 continues…