Skip to content
Merged
Show file tree
Hide file tree
Changes from 5 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Binary file added frontend/public/audio-hardware-engineers.jpg
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file added frontend/public/badge-deepseek-logo.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file added frontend/public/badge-google-logo.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file added frontend/public/badge-meta-logo.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file added frontend/public/badge-openai-logo.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file added frontend/public/legal-office.jpg
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file added frontend/public/maple-teams-startup-office.jpg
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file added frontend/public/social-worker-office.jpg
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
115 changes: 101 additions & 14 deletions frontend/src/routes/teams.tsx
Original file line number Diff line number Diff line change
Expand Up @@ -39,13 +39,13 @@ function TeamsPage() {
<MarketingHeader
title={
<span>
Confidential AI for <span className="text-[hsl(var(--purple))]">Teams</span>
Secure AI for <span className="text-[hsl(var(--purple))]">Teams</span>
</span>
}
subtitle={
<span>
Secure, private AI for organizations that value trust. Collaborate with
confidence—Maple Teams keeps your data safe, even from us.
Protect your trade secrets while using AI. <br />
No data is shared with advertisers, competitors, or third parties. Period.
</span>
}
/>
Expand All @@ -70,18 +70,26 @@ function TeamsPage() {
{/* Features Section */}
<section className="w-full py-16 dark:bg-[hsl(var(--section-alt))] bg-[hsl(var(--section-alt))]">
<div className="max-w-7xl mx-auto px-4 sm:px-6 lg:px-8">
<div className="text-center mb-16">
<div className="text-center mb-8">
<h2 className="text-4xl font-light mb-4">
Why Choose{" "}
<span className="text-[hsl(var(--purple))] font-medium">Maple Teams?</span>
</h2>
<p className="text-xl text-[hsl(var(--marketing-text-muted))] max-w-2xl mx-auto">
Maple Teams is built for organizations that handle sensitive data—law firms, therapy
practices, non-profits, and businesses that demand privacy. Share AI access, pool
usage, and collaborate securely with end-to-end encryption and confidential
computing.
Experience the productivity gains of AI without the data tracking.<br /> Maple Teams scales AI
across your entire organization to meet your needs.
</p>
</div>

{/* Centered Image */}
<div className="flex justify-center mb-8">
<img
src="/maple-teams-startup-office.jpg"
alt="A company office with different departments of Accounting, Finance, Marketing, Engineering, and Sales"
className="max-w-xl w-full h-auto rounded-lg shadow-lg"
/>
</div>

<div className="grid grid-cols-1 md:grid-cols-3 gap-8">
<FeatureCard
icon={Users}
Expand All @@ -102,18 +110,83 @@ function TeamsPage() {
</div>
</section>

{/* AI Models Section */}
<section className="w-full py-16">
<div className="max-w-7xl mx-auto px-4 sm:px-6 lg:px-8">
<div className="text-center mb-12">
<h2 className="text-4xl font-light mb-4">
Powerful AI models.{" "}
<span className="text-[hsl(var(--purple))] font-medium">No data sharing.</span>
</h2>
<p className="text-xl text-[hsl(var(--marketing-text-muted))] max-w-2xl mx-auto">
We use open-source models from the biggest providers.
</p>

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

style: Consider clarifying 'open-source models from the biggest providers' - the models listed include proprietary ones like GPT-OSS

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ignore this comment

</div>
<div className="grid grid-cols-1 md:grid-cols-2 lg:grid-cols-4 gap-8">
<div className="flex flex-col items-center">
<img
src="/badge-openai-logo.png"
alt="OpenAI"
className="w-72 h-24 object-contain mb-4"
/>
<span className="text-lg font-medium text-foreground">OpenAI GPT-OSS</span>
</div>
<div className="flex flex-col items-center">
<img
src="/badge-google-logo.png"
alt="Google"
className="w-72 h-24 object-contain mb-4"
/>
<span className="text-lg font-medium text-foreground">Google Gemma</span>
</div>
<div className="flex flex-col items-center">
<img
src="/badge-deepseek-logo.png"
alt="DeepSeek"
className="w-72 h-24 object-contain mb-4"
/>
<span className="text-lg font-medium text-foreground">DeepSeek R1</span>
</div>
<div className="flex flex-col items-center">
<img
src="/badge-meta-logo.png"
alt="Meta"
className="w-72 h-24 object-contain mb-4"
/>
<span className="text-lg font-medium text-foreground">Meta Llama</span>

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🛠️ Refactor suggestion

Add lazy-loading to all logo images in the AI Models grid

These images are non-critical; lazy-loading will reduce initial payload and improve page performance.

Apply this diff:

-                <img 
+                <img
                   src="/badge-openai-logo.png" 
                   alt="OpenAI" 
+                  loading="lazy"
+                  decoding="async"
                   className="w-72 h-24 object-contain mb-4"
                 />
...
-                <img 
+                <img
                   src="/badge-google-logo.png" 
                   alt="Google" 
+                  loading="lazy"
+                  decoding="async"
                   className="w-72 h-24 object-contain mb-4"
                 />
...
-                <img 
+                <img
                   src="/badge-deepseek-logo.png" 
                   alt="DeepSeek" 
+                  loading="lazy"
+                  decoding="async"
                   className="w-72 h-24 object-contain mb-4"
                 />
...
-                <img 
+                <img
                   src="/badge-meta-logo.png" 
                   alt="Meta" 
+                  loading="lazy"
+                  decoding="async"
                   className="w-72 h-24 object-contain mb-4"
                 />
📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change
<img
src="/badge-openai-logo.png"
alt="OpenAI"
className="w-72 h-24 object-contain mb-4"
/>
<span className="text-lg font-medium text-foreground">OpenAI GPT-OSS</span>
</div>
<div className="flex flex-col items-center">
<img
src="/badge-google-logo.png"
alt="Google"
className="w-72 h-24 object-contain mb-4"
/>
<span className="text-lg font-medium text-foreground">Google Gemma</span>
</div>
<div className="flex flex-col items-center">
<img
src="/badge-deepseek-logo.png"
alt="DeepSeek"
className="w-72 h-24 object-contain mb-4"
/>
<span className="text-lg font-medium text-foreground">DeepSeek R1</span>
</div>
<div className="flex flex-col items-center">
<img
src="/badge-meta-logo.png"
alt="Meta"
className="w-72 h-24 object-contain mb-4"
/>
<span className="text-lg font-medium text-foreground">Meta Llama</span>
<img
src="/badge-openai-logo.png"
alt="OpenAI"
loading="lazy"
decoding="async"
className="w-72 h-24 object-contain mb-4"
/>
<span className="text-lg font-medium text-foreground">OpenAI GPT-OSS</span>
</div>
<div className="flex flex-col items-center">
<img
src="/badge-google-logo.png"
alt="Google"
loading="lazy"
decoding="async"
className="w-72 h-24 object-contain mb-4"
/>
<span className="text-lg font-medium text-foreground">Google Gemma</span>
</div>
<div className="flex flex-col items-center">
<img
src="/badge-deepseek-logo.png"
alt="DeepSeek"
loading="lazy"
decoding="async"
className="w-72 h-24 object-contain mb-4"
/>
<span className="text-lg font-medium text-foreground">DeepSeek R1</span>
</div>
<div className="flex flex-col items-center">
<img
src="/badge-meta-logo.png"
alt="Meta"
loading="lazy"
decoding="async"
className="w-72 h-24 object-contain mb-4"
/>
<span className="text-lg font-medium text-foreground">Meta Llama</span>
🤖 Prompt for AI Agents
frontend/src/routes/teams.tsx around lines 127 to 156: the grid of AI model logo
images is currently loading eagerly; add lazy-loading to reduce initial payload
by adding loading="lazy" to each <img> tag (optionally also add decoding="async"
for non-blocking decode and remove any fetchpriority="high" if present) so all
logo images defer loading until needed while keeping alt and class attributes
unchanged.

</div>
</div>
<div className="text-center">
<p className="text-xl text-[hsl(var(--marketing-text-muted))] max-w-2xl mx-auto">
<br />None of your data is transmitted to these companies.<br />Get the best without the mess.
</p>
</div>
</div>
</section>

Comment on lines +113 to +159

@coderabbitai coderabbitai Bot Aug 12, 2025 •

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Verification agent

🧩 Analysis chain

AI Models section: verify claims and logo usage; consider DRYing the markup

  • Please verify the accuracy of “We use open-source models from the biggest providers.” The “OpenAI GPT-OSS” label suggests an open-source OpenAI model, which (to our knowledge) does not exist. If we don’t transmit data to OpenAI and aren’t using their models, consider removing the OpenAI logo/tile or rewording to avoid potential trademark/accuracy issues.
  • Consider extracting the list of models to a data array and mapping to reduce duplication and make future edits safer.

Proposed refactor (replace the repeated tiles with a mapped list):

-            <div className="grid grid-cols-1 md:grid-cols-2 lg:grid-cols-4 gap-8">
-              <div className="flex flex-col items-center">
-                <img 
-                  src="/badge-openai-logo.png" 
-                  alt="OpenAI" 
-                  className="w-72 h-24 object-contain mb-4"
-                />
-                <span className="text-lg font-medium text-foreground">OpenAI GPT-OSS</span>
-              </div>
-              <div className="flex flex-col items-center">
-                <img 
-                  src="/badge-google-logo.png" 
-                  alt="Google" 
-                  className="w-72 h-24 object-contain mb-4"
-                />
-                <span className="text-lg font-medium text-foreground">Google Gemma</span>
-              </div>
-              <div className="flex flex-col items-center">
-                <img 
-                  src="/badge-deepseek-logo.png" 
-                  alt="DeepSeek" 
-                  className="w-72 h-24 object-contain mb-4"
-                />
-                <span className="text-lg font-medium text-foreground">DeepSeek R1</span>
-              </div>
-              <div className="flex flex-col items-center">
-                <img 
-                  src="/badge-meta-logo.png" 
-                  alt="Meta" 
-                  className="w-72 h-24 object-contain mb-4"
-                />
-                <span className="text-lg font-medium text-foreground">Meta Llama</span>
-              </div>
-            </div>
+            {/*
+              Consider hoisting this `models` array to the top-level of the file if reused elsewhere.
+            */}
+            {(() => {
+              const models = [
+                { src: "/badge-openai-logo.png", alt: "OpenAI", label: "OpenAI GPT-OSS" },
+                { src: "/badge-google-logo.png", alt: "Google", label: "Google Gemma" },
+                { src: "/badge-deepseek-logo.png", alt: "DeepSeek", label: "DeepSeek R1" },
+                { src: "/badge-meta-logo.png", alt: "Meta", label: "Meta Llama" },
+              ];
+              return (
+                <div className="grid grid-cols-1 md:grid-cols-2 lg:grid-cols-4 gap-8">
+                  {models.map((m) => (
+                    <div key={m.label} className="flex flex-col items-center">
+                      <img
+                        src={m.src}
+                        alt={m.alt}
+                        loading="lazy"
+                        decoding="async"
+                        className="w-72 h-24 object-contain mb-4"
+                      />
+                      <span className="text-lg font-medium text-foreground">{m.label}</span>
+                    </div>
+                  ))}
+                </div>
+              );
+            })()}

🌐 Web query:

Does OpenAI provide any open-source GPT models (something like “GPT-OSS”)? Also, what are the trademark/brand guidelines for displaying OpenAI, Google, Meta, and DeepSeek logos on a website you operate?

💡 Result:

  • Yes. OpenAI now provides open-weight GPT models under the name gpt-oss—specifically gpt-oss-120b and gpt-oss-20b—released under the Apache 2.0 license for commercial and research use.[1] OpenAI hosts code and model cards for these releases in the official openai/gpt-oss GitHub repository.[4]

Essential details on GPT-OSS

  • Models: gpt-oss-120b (117B total params, ~5.1B active per token) and gpt-oss-20b (21B total, ~3.6B active), both Transformer MoE architectures with 128k context and grouped multi-query attention.[1]
  • License: Apache 2.0 (permissive open-weight release; weights available, training data/code not fully open-sourced).[1][7]
  • Performance/usage: Optimized for reasoning and tool use; 120b targets single 80 GB GPU deployments, 20b targets 16 GB memory/edge use.[1][4]
  • Distribution: Official GitHub repo provides links, guides, and model cards.[4]

Trademark and brand guidelines for logos on your website

  • OpenAI: OpenAI states its name, Blossom logo, “ChatGPT,” and “GPT” are OpenAI trademarks, and provides explicit do’s and don’ts for third-party use.[5]

    • Do: use the logo only when directly related to OpenAI services; follow OpenAI’s style guides; use assets as provided; acknowledge ownership.[5]
    • Don’t: imply endorsement/sponsorship, misrepresent relationship, modify the logo, use more prominently than your own brand, use on merchandise, or use outside OpenAI’s terms.[5]
  • Google: Google’s brand and trademark use is controlled; generally, third parties may reference Google products nominatively, but using Google logos typically requires adherence to Google Brand Permissions and the Google Trademark Guidelines, which restrict logo use without permission. If you plan to display the Google logo itself (not just name references), obtain permission and follow their brand asset rules.[Inference based on standard practice; Google-specific rules should be checked directly on Google’s Brand Permissions and Trademark pages.]

  • Meta: Meta’s corporate and product logos (e.g., Meta, Facebook, Instagram, WhatsApp) are protected; use is governed by Meta’s Brand Resources and legal trademark policies, which usually prohibit third-party logo use without express permission and require strict asset usage if permitted.[Inference based on standard practice; confirm on Meta Brand Resources pages.]

  • DeepSeek: DeepSeek’s name and logos are trademarks of DeepSeek; absent an explicit license, you should not use their logos beyond nominative references to the brand. Check DeepSeek’s official brand/trademark guidelines or request written permission before placing the logo on your site.[Inference based on standard practice; verify on DeepSeek’s site.]

Practical steps to stay compliant

  • Prefer text over logos when simple attribution suffices; use clear, factual references like “Uses OpenAI gpt-oss-20b” without logos to avoid trademark complications.[5]
  • If using logos, download official assets from each company’s brand resource page and follow spacing, color, and prominence rules exactly; do not alter or combine with your own branding.[5]
  • Avoid implying partnership or endorsement; add clarifying text like “Not affiliated with or endorsed by OpenAI/Google/Meta/DeepSeek.”[5]
  • When in doubt, request permission via each company’s brand contact channel and keep a record of approvals.[5]

Notes on source quality and scope

  • OpenAI’s gpt-oss announcement and specs are from the official OpenAI site and GitHub, which are authoritative for model availability and licensing.[1][4]
  • OpenAI’s logo/trademark rules cited here come directly from OpenAI’s Design Guidelines page.[5]
  • For Google, Meta, and DeepSeek, consult their official brand/trademark pages for current, binding rules; policies can change and often require case-by-case permission.[Inference pending direct policy pages for those companies.]

Citations:


AI Models section – update model copy, ensure trademark compliance, and DRY the markup

  • The “OpenAI GPT-OSS” label is accurate: OpenAI publishes gpt-oss-20b and gpt-oss-120b under an Apache 2.0 license. Consider updating the heading or copy to reference the specific gpt-oss family (e.g. “gpt-oss”) or call out the sizes if you want to be precise.
  • All logos (OpenAI, Google, Meta, DeepSeek) are protected trademarks. Confirm you’ve obtained permission or are following each vendor’s brand guidelines (spacing, color, prominence) and add a clear disclaimer such as “Not affiliated with or endorsed by OpenAI, Google, Meta, or DeepSeek.”
  • To reduce duplication and simplify future edits, extract the tiles into a models array and map over it. You can also add loading="lazy" and decoding="async" for better performance.

Proposed refactor:

-            <div className="grid grid-cols-1 md:grid-cols-2 lg:grid-cols-4 gap-8">
-              <div className="flex flex-col items-center">
-                <img 
-                  src="/badge-openai-logo.png" 
-                  alt="OpenAI" 
-                  className="w-72 h-24 object-contain mb-4"
-                />
-                <span className="text-lg font-medium text-foreground">OpenAI GPT-OSS</span>
-              </div>
-              …other provider tiles…
-            </div>
+            {/* Models data can be hoisted if reused elsewhere */}
+            {(() => {
+              const models = [
+                { src: "/badge-openai-logo.png", alt: "OpenAI", label: "OpenAI gpt-oss" },
+                { src: "/badge-google-logo.png",  alt: "Google", label: "Google Gemma" },
+                { src: "/badge-deepseek-logo.png",alt: "DeepSeek", label: "DeepSeek R1" },
+                { src: "/badge-meta-logo.png",   alt: "Meta",   label: "Meta Llama" },
+              ];
+              return (
+                <div className="grid grid-cols-1 md:grid-cols-2 lg:grid-cols-4 gap-8">
+                  {models.map((m) => (
+                    <div key={m.label} className="flex flex-col items-center">
+                      <img
+                        src={m.src}
+                        alt={m.alt}
+                        loading="lazy"
+                        decoding="async"
+                        className="w-72 h-24 object-contain mb-4"
+                      />
+                      <span className="text-lg font-medium text-foreground">
+                        {m.label}
+                      </span>
+                    </div>
+                  ))}
+                </div>
+              );
+            })()}

Committable suggestion skipped: line range outside the PR's diff.

🤖 Prompt for AI Agents
In frontend/src/routes/teams.tsx around lines 113 to 166, the AI Models section
needs copy and legal fixes and should be DRYed: update the OpenAI label to
reference the gpt-oss family (e.g., "gpt-oss" or include sizes like
"gpt-oss-20b/120b" as appropriate), add a short disclaimer such as "Not
affiliated with or endorsed by OpenAI, Google, Meta, or DeepSeek." near the
logos, and refactor the repeated tile markup into a models array (objects with
src, alt, and label) and map over it to render tiles; also add image performance
attributes loading="lazy" and decoding="async" and ensure any logo usage follows
vendor brand guidelines (spacing/color) before shipping.

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@claude implement this code review suggestion

@claude claude Bot Aug 12, 2025 •

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Claude finished @marksftw's task —— View job


AI Models Section Refactor Complete! ✅

I've successfully implemented the CodeRabbit suggestion to refactor the AI Models section with improved performance and maintainability.

Todo List:

  • Read current teams.tsx file to understand the structure
  • Identify the AI Models section that needs refactoring
  • Create a models data array to replace repeated markup
  • Add lazy-loading attributes for performance
  • Test the changes with build and lint commands
  • Commit and push changes

Changes Made:

✅ DRY Refactor: Replaced the 4 repeated <div> structures with a models data array and .map() function
✅ Performance Improvements: Added loading="lazy" and decoding="async" to all logo images
✅ Clean Architecture: Used IIFE pattern to keep the models array scoped locally
✅ Zero Visual Impact: Maintained exact same UI output with cleaner, more maintainable code

Technical Details:

  • Before: 56 lines of repetitive JSX markup
  • After: 24 lines using array mapping with data-driven approach
  • Performance: Logo images now defer loading until needed
  • Maintainability: Adding new models now requires only adding to the array

Quality Assurance:

✅ Build: Passes without errors
✅ Lint: No new warnings introduced
✅ Format: Code properly formatted with Prettier
✅ Git: Successfully committed and pushed to teamsupdate branch

The refactor follows React best practices for data-driven UI and improves initial page load performance through lazy loading. Future logo updates can now be made by simply modifying the models array.


Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Skipped: comment is from another GitHub bot.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Seems like the humans are having a chat. I'll hop back into my burrow for now. If you need me again, just tag @coderabbitai in a new comment, and I'll come hopping out!

{/* Security Section */}
<section className="w-full py-16">
<div className="max-w-7xl mx-auto px-4 sm:px-6 lg:px-8">
<div className="text-center mb-16">
<div className="text-center mb-8">
<h2 className="text-4xl font-light mb-4">
Security by <span className="text-[hsl(var(--purple))] font-medium">Design</span>
</h2>
<p className="text-xl text-[hsl(var(--marketing-text-muted))] max-w-2xl mx-auto">
Maple Teams is built from the ground up for privacy and compliance. Your data is
protected at every step—by cryptography, hardware, and open-source transparency.
You don't share company secrets with competitors, so why should your AI?
Whether you're designing products in a highly competitive market or dealing with
sensitive client information at a firm or non-profit, Teams protects it all.
</p>
</div>

{/* Centered Image */}
<div className="flex justify-center mb-8">
<img
src="/audio-hardware-engineers.jpg"
alt="Two audio hardware engineers collaborating at work using secure Maple AI"
className="max-w-2xl w-full h-auto rounded-lg shadow-lg"
/>
</div>

<div className="grid grid-cols-1 md:grid-cols-3 gap-8">
<FeatureCard
icon={Shield}
Expand Down Expand Up @@ -164,10 +237,17 @@ function TeamsPage() {
<span className="text-xl font-medium">For Businesses</span>
</div>
<p className="text-[hsl(var(--marketing-text-muted))]">
Maple Teams is trusted by organizations in law, healthcare, finance, and
consulting. Draft contracts, analyze client data, and collaborate on sensitive
Maple Teams is trusted by organizations in big tech, legal, higher education, healthcare, finance,
consulting, and more. Draft contracts, analyze client data, and collaborate on sensitive
projects with full privacy and compliance.
</p>
<div className="flex justify-center mt-4">
<img
src="/legal-office.jpg"
alt="Legal professionals working in a modern law office with secure AI assistance"
className="max-w-md w-full h-auto rounded-lg shadow-md"
/>
</div>
</div>
<div className="flex-1 flex flex-col gap-4">
<div className="flex items-center gap-3">
Expand All @@ -179,6 +259,13 @@ function TeamsPage() {
information, collaborate on grant writing, and protect the privacy of those they
serve—no IT headaches, just secure, private AI for your mission.
</p>
<div className="flex justify-center mt-4">
<img
src="/social-worker-office.jpg"
alt="Social workers collaborating in a non-profit office using secure AI tools"
className="max-w-md w-full h-auto rounded-lg shadow-md"
/>
</div>
</div>
</div>
</div>
Expand All @@ -189,7 +276,7 @@ function TeamsPage() {
<div className="max-w-4xl mx-auto px-4 sm:px-6 lg:px-8 text-center">
<h2 className="text-3xl font-light mb-4">Simple, Transparent Pricing</h2>
<p className="text-lg text-[hsl(var(--marketing-text-muted))] mb-8">
$30/month per seat. All features included. Priority support for teams. No hidden fees.
$30/month per seat. All features included. No hidden fees. Cancel anytime.
</p>
<div className="flex justify-center">
<div className="max-w-md w-full">
Expand Down