We put the bundle.social logo on a race car and from what I’ve heard so far our driver took P2 in quali

I’m on vacation so updates are reaching me a bit late but I couldn’t be happier

Both of us are into racing and cars. We do sim racing and I even build cars in my spare time, so this is pretty personal and I will MILK everything out of this. No elaborate marketing strategy here, just something we’ve always wanted to be part of

Pretty surreal seeing our logo on an actual race car. More pro pics coming soon!

Car went through

3 engines

3 gearboxes

2 paint jobs

But at last, finally finished; just gotta put some better tires

We originally hired someone on Fiverr to create a product animation for bundle.social, the social media API we’ve been building for the past few years.

Then they stopped replying.

So instead of waiting, we tried building the first version ourselves. The video attached to this post was created from a self-contained HTML file in roughly 40 minutes, including a tea break.

FYI, RECORDING IS A BIT WONKY BECAUSE I EXTRACTED IT FRAME BY FRAME

The process was basically:

1. We asked AI to break down the advertising styles of Apple, Solana and Revolut into actual rules: shot lengths, transitions, easing, text movement and end-frame timing.
2. We used that document as the specification for a bundle.social animation.
3. We slowed everything down, fixed the loop, and stopped the model from “improving” our product claims.

The final result is a 35-second animation with six scenes. It borrows Solana’s fast text, Revolut’s movement and CTA structure, and Apple’s typography and long holds.

Is it animator-level work? Not really. It got us about 70% of the way there. There’s no sound design, real footage, 3D, or that final layer of polish.

But instead of giving an animator a mood board and saying “make it feel modern,” we can now give them a working reference with exact timing, copy, structure, and feeling.

If you want to see the full process, including the actual prompts, what broke and where this method stops being useful, we wrote everything down here:

https://bundle.social/blog/product-demo-animation-three-prompts

Do you guys like it as an MVP?

Noticed a small update inside the Meta App Review dashboard that is actually a pretty big deal if you build anything using Facebook, Instagram, Messenger, or WhatsApp APIs. We do so, that's why this and our blog post exist.

It now says:

>Most submissions are reviewed within 20 days. If we need further information or ask you to resolve issues, the process could take longer.

Not long ago (about a week ago), the same dashboard said 12 days.

I’ve been through like 10 separate Meta App Reviews over the years (admittedly not a lot, but everything has worked out so...) and the timeline has basically evolved like this.

In 2024, it was the same day, quick and easy, you could bang them out in 1h intervals, then

• 2025: usually 1 - 3 days
• Jan-Feb 2026: around 4 - 6 days
• March-April: 9 - 12 days
• May: 17- 19 days
• Now: Meta officially tells you to expect up to 20 days

So what’s the issue?

The issue is, if or rather, when you mess up. You wait 20 days, get rejected because the test account does not work, the reviewer hits a paywall, or your Loom video skips the exact permission flow they wanted to see.

Then you fix it, resubmit, and go straight back to the end of the queue. Congrats, your two-week integration is now a two-month journey. For us, it was 43 times when they were scrutinizing things really harshly.

Why the slowdown? My best guesses:

• way more people shipping apps with AI
• half-working integrations are being submitted for review
• developers requesting every permission “for later”
• vague descriptions like “we use this to improve user experience”
• reviewers having absolutely no time to debug your product for you

App Review is not your QA department.

The reviewer wants to see the exact permission being used inside the exact UI you described, with working credentials, without watching 11 minutes of you randomly clicking.

What has worked best for us:

1. Request only the permissions the app uses right now.
2. Start the video from the login screen.
3. Show the complete OAuth flow.
4. Trigger the exact feature connected to the permission.
5. Show the final result on a real test account.
6. Test everything in Incognito before submitting.
7. Make the review notes painfully literal.

Make the review notes painfully literal.

Oh, and make sure the test account is active and actually works.

You give the reviewer login credentials to your app. They are not going to create a new account, configure everything, connect Meta, and unlock the right features just to test your submission.

If your app has premium features, every feature shown in the review video needs to be enabled on that test account too.

The reviewer should be able to log in, follow your instructions, click the exact buttons you described, and see the exact result you promised.

Bad:

>Log in and test the publishing feature.

Better:

>Log in using the credentials provided below. Open the “Create Post” page from the left sidebar, select the connected Instagram account, upload an image, enter a caption, and click “Publish Now.” The premium publishing feature is already enabled on this account.

Are you also seeing 20-day reviews now, or do we have a skill issue?

those are real listings from my city

each around $400k btw

I’m currently on a “search” for a large apartment to buy, which basically means I doomscroll real estate listings.

So I scroll past this flat: $400k for 136m². okay, a bit much on both price and size, but fine

Maybe it’s one of those old apartments with great bones, bad furniture. The title says investment property so I already know what game we’re playing

I’m expecting “needs renovation" or “rough around the edges.” Something that I could do myself + a cold one

Then I open the photos.

• First image: beautiful AI-generated dream flat.
• Second image: another perfect render

You get the idea, and then, at the very end, there’s one actual photo.

bruh

The place looks like it was bombed out during the Chechen wars, abandoned and briefly used as a drug den.

Everywhere else, when you buy something, the listing is at least vaguely connected to reality.

You search for a car, you see the car. Maybe it’s washed, maybe the photo angle hides a scratch. But still, it’s a car.

• You search for clothes, you see the clothes.
• You search for a phone, you see the phone.

But real estate? Bruh, at least I should get a pop-up while clicking on a listing to tell me that I should wear lipstick because I should be looking nice before I'm getting F-ed

If this was illegal, or at least illegal in Poland, I genuinely think prices for destroyed apartments would drop back to reality overnight. Because no sane person would pay premium money for a place if the listing had to be honest.

Just:

>Shack in the middle of the city
Tree growing inside
150k

And im cool, will have a nice Japanese garden inside my new property

Being honest, I was very sceptical about adding trial tiers. I liked it when people were messaging us on Chat asking for a trial or a free dev month; there was a human aspect to it, we had meetings, networked, and sometimes made friends.

We wanted to try sth new, added the trias, and in a matter of weeks we had like 10/30 new customers a day. (The 30 is a bit of a stretch because it was one time, but you know $$$$)

Why is it the worst then?

Look at the screenshot, all the people there used the service and then just left, or worse, liked the service but wanted to game the system, created multiple trials, and were hopping from one to the next. Luckily, after turning off the service for them in the middle of their night 3 times, they decided to convert into paying ones and had a bit of a laugh

Do you have any thoughts on trials or maybe methods to stop abusers on your platform?

I was on the market for my first gun and had the budget to afford the Shadow 2 + IPSC accessories like belts/holsters.

CZ decided to do a special edition for the 35-year anniversary of “Hunters Day” in Poland. Every shop in Poland had like 2 or 3, and they were sold out everywhere.

I called around almost the whole country looking for one, and at last I called my local small hunting shop. They had one in stock, but they weren’t doing reservations, so I ran to get it.

When I entered the shop, two other dudes entered right behind me. When they learned that I was the last one to get it, they started some friendly banter, saying they should’ve run when they saw me running earlier.

Price: 900€ / $1000

Downside: 2022 model (but new)

Toni's system grips and magwells: acquired

Yooo,

The title is a little bit of clickbait because my actual intention was more along the lines of "vibe coders struggled, so we helped them." This is supposed to be a tutorial post.

Why?

Because a lot of people building things today aren't programmers at heart, and the struggle is real. We get hundreds of support messages every day for "bugs" that turn out to be caused by inexperience, incorrect assumptions, or simply not reading the docs.

After seeing the same pattern over and over, the solution became pretty obvious: add an MCP server and a CLI so Claude can check your work before you get frustrated at us.

So here's a small tutorial.

Claude is already pretty good at writing social posts, building dashboards, and generally figuring things out. Give it a brand_voice.xml file or the shape of your data, and you're already halfway there.

What bundle.social MCP actually does

The MCP server exposes our API as tools Claude can use.

• check if your API setup works
• list connected social accounts
• create/schedule/retry/fetch posts
• fetch post analytics
• fetch social account analytics
• import post history
• create comments
• bulk schedule from CSV
• create hosted portal links so users can connect their own accounts
• work with teams/workspaces

There are also platform helper tools for annoying stuff like

• subreddit requirements / flairs
• YouTube categories / playlists / regions
• LinkedIn mentions
• Google Business (like everything you need)
• TikTok trending music

There are a lot of things you can do, all available in the docs section on our landing.

Setup

The MCP server runs locally through npx.

For Claude Code:

claude mcp add bundlesocial \
  --env BUNDLESOCIAL_API_KEY=sk_live_... \
  -- npx -y bundlesocial-mcp

For Claude Desktop, add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "bundlesocial": {
      "command": "npx",
      "args": ["-y", "bundlesocial-mcp"],
      "env": {
        "BUNDLESOCIAL_API_KEY": "sk_live_..."
      }
    }
  }
}

If you use multiple teams/workspaces, you can also set:

BUNDLESOCIAL_TEAM_ID=team_...

Docs are here:

https://info.bundle.social/api-reference/mcp

The basic workflow

The main thing I would not do is this:

Post this everywhere.

That is how you get garbagio. The better pattern is to make Claude inspect first.

Example:

Check my bundle.social setup and list connected social accounts.

Tell me which accounts can publish posts and which ones need extra channel selection.

Do not create or schedule anything yet.

This matters because “social account” is rarely just one simple thing.

For example:

• Facebook usually needs a Page
• LinkedIn can be a personal profile or an organization
• Discord needs a server/channel
• Pinterest needs a board
• Reddit may need subreddit requirements and flair
• YouTube has categories, playlists and regions

So Claude should check the real setup before trying to create anything.

Otherwise, it starts guessing, and guessing + publishing tools = Slop top

Drafting platform-specific posts

Once Claude knows what accounts are connected, you can ask it to draft different versions.

Example:

We shipped MCP support for bundle.social.

Draft platform-specific posts for LinkedIn, X, Bluesky and Reddit.

LinkedIn should be more detailed and developer-focused.
X should be short.
Bluesky can be more casual.
Reddit should be tutorial-style and not sound like a press release.

Do not create or schedule anything yet.

Claude is already useful here.

The difference is that now the next step can happen in the same workflow instead of becoming manual copy-paste hell.

Approval before scheduling

After Claude gives you the drafts, you review them and approve one version.

Example:

Use version 2.

Schedule the LinkedIn post for tomorrow at 9:00 UTC.
Schedule the Instagram post for tomorrow at 9:10 UTC.
Schedule the Bluesky post for tomorrow at 9:20 UTC.

Create the Reddit post as a draft only.

Return all post IDs, platforms and scheduled times.

This gives you a pretty sane human-in-the-loop flow:

1. Claude checks the setup.
2. Claude drafts the content.
3. Human approves the final version.
4. Claude creates/schedules the posts.
5. The API handles publishing.
6. Claude can later check analytics.

So you whip Claude for the boring API work, but it does not get to YOLO, WE BALLIN your company account.

(which is probably healthy)

Media uploads

Media was one of the bigger reasons we wanted this.

A lot of AI social workflows are fake-useful until you need to attach an actual image or video, with the MCP server, Claude can upload media first and then use the uploaded media in a post.

Example:

Upload ./demo-video.mp4.

Prepare a LinkedIn post and an X post using that video.

Show me the final copy and media attachment summary before scheduling anything.

Then, after review:

Looks good.

Schedule LinkedIn for tomorrow at 10:00 UTC and X for 10:15 UTC.

Return the created post IDs.

That is much nicer than Claude saying:

>“Attach your video here.”

Thanks bro, very helpful.

Analytics loop

The first post is not even the most interesting part.

The more interesting part is what happens after publishing.

Because once posts exist in bundle.social, Claude can fetch analytics later and use them as context.

Example:

Check analytics for the posts we scheduled last week.

Group results by platform.

Tell me which topic performed best, which format underperformed and what we should test next week.

Or:

Pull analytics for this post: post_abc123.

Explain why it may have performed better than our previous launch post.

Draft a follow-up post for LinkedIn and X.

Why the “no account limits” thing matters

This is the part people glaze us for the most.

A lot of social media tools are priced around seats or connected accounts. I get it, it PRINTS money, but it's terrible for clients, and if you are building for

• agencies
• SaaS products
• franchises
• multi-location brands
• creator networks
• internal tools
• products where end users connect their own social accounts

uuuuuuhhh you aaaare getting rinced.

We do not do artificial connected account limits, usage limits and official platform limits still exist, obviously.

Meta is still Meta. TikTok is still TikTok.

But connected accounts are not the thing we cap, which makes MCP workflows nicer because Claude can work across a real multi-account setup instead of one hardcoded social profile.

Example full prompt

Check my bundle.social setup and list connected social accounts.

Then draft a launch announcement for our new MCP server.

Create separate versions for LinkedIn, Bluesky and IG.

For IG, make it tutorial-style with graphics.
For LinkedIn, explain the developer workflow in more detail.
For Bluesky, make it more casual.

Do not create or schedule anything until I approve.

Then after approval:

Use the approved versions.

Schedule LinkedIn for tomorrow at 9:00 UTC.
Schedule X for 9:10 UTC.
Schedule Bluesky for 9:20 UTC.

Create the Reddit post as a draft.

Return all post IDs, platforms and scheduled times.

Package:

npx bundlesocial-mcp

Docs:

https://info.bundle.social/api-reference/mcp

Also, if you are a data nerd, we have raw unparsed analytics: https://info.bundle.social/api-reference/platforms/instagram#raw-analytics-demographics-&-audience-data

We are at a point where we get a lot of clients daily for our social media API, but I still refuse to hire someone for support, as I find it funny/enjoyable most of the time. Today's email was the funniest thing this month.

Hope the boys will respond, and we can have a corporate group chat and we can talk about "stuff"

Hi, we built a dashboard for managing unlimited locations after an SEO agency asked us for it. They are using it, but they do not give much feedback, so I’m trying to get a second opinion. I added screenshots of the main panels, so no need for a link. Im mainly interested in:

Does this layout make sense for SEO work? Or the flow is meh?

Anything you would move, remove, or make more visible?

For context we are mostly API-first, but I still want the dashboard to be usable for the non-IT crowd.

The last screenshot shows the team switcher and the empty space on a monitor (granted it's ultrawide, but you know there is a room for activities)

Alright, small update for anyone building Instagram/Reels publishing workflows.

We’ve had TikTok music library support for a while, and Meta now has something similar on the Instagram side.

You can search for:

• pre-cleared music
• user-posted cleared music
• a specific sound directly, if you already have the sound ID

The screenshot is from our UI, but the main idea is pretty simple: pick the track, attach it to the post, and set the audio levels.

The “music volume” and “original audio volume” fields are just controls for how loud each audio source should be.

The annoying catch, because of course there is one:

This only works when the Instagram account is connected through Facebook Business Login. It does not work with the standard Instagram connection flow.

So if you are wondering why music search/attachment works for one IG account but not another, that is probably the reason.

Tiny plug for context: we run bundle.social, a social media API/dashboard with no account limits

Howdy all, Im Marcel, and for some time now, we have had a TikTok music API in our system that is used by our content distribution customers to promote new music for artists, so I thought we should not gatekeep that.

Being upfront, we don't do sh**t to promote you. We are a postman for your service.

What we do is sth like this

An artist has a song or sound they want to push.
Fans connect their own TikTok accounts through OAuth.
Your app prepares the post flow with the chosen music.
The user publishes the TikTok from their own account with the selected sound.

That gives artists, labels, agencies, or music marketing tools a way to build organic TikTok campaigns with data to go off as we also provide analytics for those posts like watch time and audience splits.

The key is to have your sound/music already on TikTok, I mean, you don't gotta have it, but it won't be that nicely tagged.

Curious if anyone here is running fan/creator-based TikTok campaigns for artists, or im in complete wrong place.

I've been preaching a lot about lean development and building things on request for your clients, but I failed to mention that sometimes it can backfire.
So what happened?

One of our clients decided that he no longer needs our services as a social media API but would love to stay if we could provide a Google Business Profile API. He wanted the whole thing, meaning reviews, business hours, special hours, attributes, categories, media, services, food menus......

tbh it made sense from a product standpoint Google Business Profile is one of those things that sits in the weird corner between SEO and social media

So we built it, tested it. Packaged it nicely into the UI and in the API

Then I sent an update email to subscribers basically saying:

>yooo we got it like that go crazy

Client replies almost instantly.

I’m thinking, perfect, lean development all that.

Instead I get:

>Good afternoon Marcel, you're going to kill me but I have to cancel

10/10 absolute cinema.

To be fair, I’m not mad at him. We are still in touch. Things change. Budgets change. Priorities change.

People say “we would stay if…” and sometimes they actually mean it at that exact moment, but the exact moment expires before your sprint does.

So yeah.

Build close to customers, but remember things change as fast for them as for you.

How you guys handle situations like those

and for the plug:

- we added a 14 day trial to our social media API, so if you wanna check out the GBP integration, go crazy

Howdy, I’m Marcel, one of the nerds behind a social media / local business API product.

Small disclosure: this is kinda a plug as I build it. Sooooooo

I recently added fuller Google Business Profile support into our dashboard and API. The idea is to make GBP easier to manage alongside other channels instead of treating it like a separate, weird little island.

Right now the focus is on things like:

• reviews
• business hours, including special hours etc....
• menus
• categories
• attributes
• services
• media

The reason we built it is pretty simple if you manage more than one location and have changing hours, seasonal services, it's kinda pain in the a** to run + you get the rest of social media scheduler/API included.

Even if you are not interested in testing it out, what did I miss that seems obvious?