Your voice at the center.
Every channel in orbit.
You want to be building the next idea, not staring at a blank posting box. Going quiet cost me 77% of my reach.
Today that means X and Instagram. Your blog, TikTok, YouTube, newsletters and RSS are next.
- Connects from Claude Code, Claude, Codex, ChatGPT.
- X: live
- Blog: coming soon
- Instagram: live
- TikTok: coming soon
- YouTube: coming soon
- Newsletter: coming soon
- RSS: coming soon
claude and openai in, seven channels out, two lit today
It runs where you
already work.
No second app to keep open, no dashboard to remember. You are already talking to Claude about the thing you shipped. This is that conversation, with vibedraft connected to it.
can you craft 6 posts for vibedraft_app spread out for aug 22 saturday. understand how it performs on saturdays and try and replicate structure on what performed
Used vibedraft integrationget_post_metrics100 postsAug 15 is the only full Saturday on record. The honest question at 8:45pm took the reach, the numbered listicle took the bookmarks, and the dogfooding proof at 11:30am took the engagement. Six drafts, mirrored to those slots.
yes schedule them
Used vibedraft integrationschedule_post× 6All six queued for @vibedraft_app. 6 credits charged.etpost11:30amdogfooding proof1:30pm“where vibedraft is going” listicle3:02pmbest marketing hire4:49pmfive models listicle6:46pmsignals / data lake8:43pmhonest questionaccountvi
@vibedraft_appcost 1 credit each, on your yesThe building part was
never the problem.
Founders on X, describing the job that starts after the product works. None of them are customers, none of them were asked, and every quote links to the post it came from.
is anyone else sick and tired of marketing? i just want to build
jb
@Jbm_dev1 AprYou can vibe-code an MVP in days, but getting it in front of the right users is a completely different challenge.
dh
@Dhruvam98712 JunMarketing is at least 10x harder than building. Maybe 100x. That's why so many great products never take off.
th
@thenowhereway14 JulI have so many ideas. But not near enough time to make them all! does this happen just to me?
ja
@Jacob_Rhodes_20 Junshipping 10 products with no distribution is worse than shipping 1 with a real audience.
ar
@aryanlabde25 Maywhat should i post today? i ask myself that every single day but it feels like i have lost my posting mojo lately
jb@Jbm_dev15 JunDistribution isn't a growth strategy. It's the business. Without it, nobody knows you exist.
1u
@1Umairshaikh8 JunJust curious, how do you manage to post more than 3–4 times a day? I've been trying for the last few weeks, but I always stop after 3 posts
ak
@Akasheth_10 JulEvery founder now has two jobs: 1) building the product 2) building the audience. How do people get users with no audience?
1u@1Umairshaikh1 Junbuilding a SaaS solo is wild. you handle: the product, the design, the marketing, the support, the bugs, the refunds. all at once.
ar@aryanlabde9 JulThe hardest part of building in public isn't the vulnerability. It's posting when there's nothing exciting to report.
ji
@Jia_builder22 Mayevery vibe coder says "I'll do marketing after launch." launch comes. marketing never starts.
1u@1Umairshaikh1 Julthought i’d post 5 times a day but ended up posting 1-2 times
ak@Akasheth_12 AugFour jobs. Three of them
already run without you.
Growing an account is a loop. Decide what is worth saying, write it, ship it on a cadence, then read what it did and let that change the next one. vibedraft closes three of those four today. The fourth is greyed out here for the same reason the unbuilt channels are: it does not exist yet.
What is worth saying this week, given what you just shipped and what the numbers are doing. Today this part is you, in one sentence.
Drafts in your voice, built from the last 500 posts on whichever account it is writing for.
get_voice_contextQueued into slots that have worked for that audience, and posted through the official X API once you say yes.
schedule_postImpressions, replies, bookmarks and follows come back per post, and the next draft is written against them.
get_post_metricsIt never publishes without you.
Autonomous drafting is the point. Autonomous publishing is how founders get burned. vibedraft holds the line between the two, permanently.
Your yes is the gate
Every post waits for an explicit approval, in the chat or on the board. There is no setting that turns this off.
OAuth only, never a password
You connect through X’s own authorization flow. vibedraft never sees a password, and you can revoke access from X settings in one click.
Built on the official X API
Browser session bots violate X terms and get accounts suspended. vibedraft is the pipeline agents are actually allowed to use.
Set up in under two minutes.
One command, one authorization, one sentence. There is no dashboard to learn before you get value, and nothing to configure before the first draft lands.
claude mcp add --transport http vibedraft https://vibedraft.app/api/mcpclaude/mcphttps://vibedraft.app/api/mcp[mcp_servers.vibedraft]
url = "https://vibedraft.app/api/mcp"
auth = "oauth"codex mcp login vibedrafthttps://vibedraft.app/api/mcpThree products.
One of you.
Every founder account dies the same way. A launch week eats the calendar, posting stops for six weeks, and the algorithm quietly forgets you exist. Reach is not talent. It is a habit, and habits break under load.
My posting days ran 4.3x the reach of my quiet ones. This is what going quiet cost my own account. Measured across 1,728 posts on @rcmisk between 15 April and 30 July 2026, days grouped by how much I posted and impressions read from the highest snapshot of each post. Not a case study.
Ask for a week. Get a week.
You describe the week in the chat you already have open. Claude writes each post against the right account’s voice profile, spaces them into slots that have worked for that audience, and queues them.
Ask your numbers a question.
Every post reports back. The same connection that drafts and schedules also reads impressions, likes, replies, reposts, bookmarks, follows gained, and profile visits, per post, on every account you run. You query it in the chat you already have open, in plain language.
Real response from @rcmisk, 18 August 2026. WES is the weighted engagement score, returned per post.
The read-only tools your agent already has
get_post_metricsImpressions, likes, replies, reposts, bookmarks, follows gained, profile visits and a weighted score, per post. Filter with since, cap with limit.
get_signalsWhat is measurably working on this account right now, written as a brief your agent drafts against. The one call here that costs credits.
list_scheduledEverything queued, with status pending, posted, failed or canceled, and the tweet id once it ships.
list_accountsEvery handle this login drives, and which one the connector is pointed at right now.
get_daily_packVoice profile, live signals and the current queue in a single call, for the start of a drafting session.
top_postsYour best performing posts, optionally per weekday. Real exemplars to draft from, not a description of your style.
growth_postsThe posts that drove people to your profile, ranked by rate rather than reach.
engagement_by_local_hourAverage engagement by hour of day, so a post lands where this account actually performs.
engagement_by_dow_and_hourThe same, split per day of the week, with thin days flagged rather than guessed at.
search_my_postsSearch your own published posts by keyword, so a draft does not repeat something you already said.
One connection. Every surface you work in.
vibedraft speaks MCP, so claude.ai, Claude Code, ChatGPT and Codex all drive it from the window you already have open. Connecting an account reads its last 500 posts and builds a voice profile for it. There is no dashboard to learn before you get value.
One connection in. Three surfaces out.
Anything that speaks MCP can hold the connection: the server publishes both discovery documents and registers clients dynamically, so it is one OAuth either side of the fence. What is greyed out has no tool behind it yet, so it is not sitting on the page pretending otherwise.
publish_postpublish_articlepublish_blogget_post_metrics · get_signalsEverything the connector can do.
Every card is a real call on the vibedraft MCP, reachable from the chat window you already have open.
There is a full app too, for the days you would rather click than type. The same queue, the same accounts, the same voice profiles.
One board. Every audience.
Each account keeps its own voice, cadence, and slots. You read one board. Edit inline, drag to move, approve a whole row at once.
| Account | Monday | Tuesday | Wednesday |
|---|---|---|---|
rc @rcmiskFounder voice | 9:00 AMNobody warns you that shipping is lonely.posted | 1:00 PMThe graph I wish I had seen in week two.queued | 1:00 PMWhat a content pipeline actually has to do.needs your yes |
vd @vibedraftProduct | 11:30 AMYour agent can write and post for you now.posted | 4:30 PMOne MCP command. That is the whole setup.queued | 4:30 PMWhy we will never auto publish.queued |
tt @trusttrafficProduct | 8:15 AMLaunch day. Verified traffic badges are live.queued | 9:15 AMDashboards give you a number, not belief.needs your yes | 9:15 AMThree founders, three badges, one week.needs your yes |
Approved posts fire from a dispatcher that runs every 5 minutes. Illustrative board, real product behavior.
It writes like you, then learns what worked.
A profile per account, not one house style
Generic AI writes for everyone, which is why it sounds like no one. vibedraft generates against a profile derived from each account's own posts: tone, length habits, opener patterns with counts, and the phrases that account never uses. It is plain markdown and you can read every line of it. Your product account can sound nothing like you, on purpose.
tone irreverent and direct, allergic to
corporate cliches and false optimism
avoids synergy, growth hacking, move the
needle, best practices, emojis
length mostly 1 to 3 sentences
source the account's own posts, last
rebuilt 13 Aug 2026
Results route back into the next batch
The measured half of the profile is not tagged by hand. Each topic cluster carries the engagement level it actually earned, each opener shape carries how many times it was observed, and every exemplar is one of your own posts with its real numbers attached. The top one currently reads 7,412 impressions, 125 likes, 127 replies.
▲ distribution as bottleneck very high
▲ build versus distribute very high
▲ cadence and compounding high
▼ landing pages and funnels low
→ a topic that stops landing drops off
Read it, change one line, put it back.
The profile is yours to overwrite. Two routes, and they touch the same markdown.
In the app
Open Voice profile and edit the markdown directly, section by section. The same screen shows which exemplars are currently anchoring the model, so you can drop one that no longer sounds like you.
In the chat
get_voice_context returns the exact markdown that update_voice_profile overwrites, so your agent can read it, change one section, and write it back without disturbing the rest.
read my voice profile and soften the tone sectionWhat moves on a clock, and what waits for you
Two of the three refresh themselves. The one that decides how you sound does not, and that is deliberate.
| Your numbers | every 2 hours | Impressions, likes, replies and reposts for anything posted in the last 14 days. |
| Your signal brief | every morning | What is measurably working right now, recomputed daily over a 7 to 30 day window. |
| Your voice profile | when you ask | Built when you connect the account, rebuilt when you press regenerate, changed when you edit it. Never behind your back. |
Some rules the profile cannot override
Style is yours. These are not style. An em dash is rewritten for you before you see the draft. The rest bounce at preview, before anything reaches your queue.
Genuine questions are the exception, and they are encouraged. A question that takes a real thought to answer is the strongest move a small account has. One that takes a tap is bait.
no reaction bait, the “thoughts?” kind
no viral templates
no LinkedIn slop
no “as an AI” meta
nothing over 4,000 characters
→ then scored out of 5 by the drafting model, aim 4
Start with a 7 day free trial.
Cancel before it ends and you are not charged. Full refund in the first 30 days after that, no questions. You are a builder, and posting daily should not require becoming a marketer.
Basic
- 1 X account
- 1,000 credits a month, roughly 30 posts a day
- 100 in app generations a month
Pro
- 3 X accounts, each with its own voice profile
- 3,000 credits a month across all accounts
- 300 in app generations a month
- A personal account, a product account, and room for the next one
In both plans
- Claude MCP connector with uncapped drafting
- Voice profile per account from your last 500 posts
- Conversational composer for planning a whole batch
- Multi account board with drag to reschedule and bulk approve
- Dispatcher that posts approved work every 5 minutes
- Engagement tracking you can ask questions of in chat
What a credit buys
| A plain post | 1 credit |
| A post carrying a link | 13 credits |
| A signal bundle | 1 credit |
| Reading your own metrics | free |
Links cost 13 credits because X charges roughly thirteen times more to distribute them. In my own data the link only post took 6 impressions while the same idea written out took 40, so the cheap option is usually the better one anyway. Asking your numbers a question never costs anything.
Things you are probably thinking.
How is this different from Typefully or Hypefury?
Those are schedulers. They solve posting, not writing, and they know nothing about your audience. vibedraft drafts in each account’s voice, learns from real engagement, and plugs into Claude natively over MCP. It is a pipeline, not a queue.
Will it sound like me, or like AI slop?
Every draft is generated against a profile built from that account’s last 500 real posts, including the patterns it never uses. If a draft misses, edit it inline or ask for a rewrite. Nothing posts until you approve it.
I run more than one product. Does that actually work?
That is the design target. Each account carries its own voice profile, its own slots, and its own performance history. One chat plans across all of them, and one board approves them.
I have 30 minutes a day. Is that enough?
That is the constraint the product was built around. One chat batches the week, one pass over the board approves it. Most of the work happens where you already are, inside Claude.
What is a credit, and will I run out?
A plain post costs 1 credit, so Basic’s 1,000 a month is more than 30 posts a day. A post carrying a link costs 13, because X prices link distribution about thirteen times higher. Reading your metrics is free.
Your audience compounds while you keep shipping.
7 day trial, then $19.99 a month. Cancel any time from the app.