On this page9 sections
Why we are building Munshi
Every small business we know has the same gap. Marketing matters, nobody owns it, and the tools all assume someone will.
The bet behind Munshi Social, the AI content marketing engine for small teams, is that the missing piece is an operator, not another tool. It plans, writes, publishes and reads results as one loop, and a person supervises it in a single weekly sitting.
Munshi runs its own marketing
The articles on this site come from the same engine customers pay for, and a person approves them like everything else. When the output embarrasses us, we feel it before any customer does.
What shipped in the last few weeks
These are the biggest changes from the last few weeks.
- A scheduled post that cannot be sent within two hours of its time is now held back. It asks for a new time instead of going out late, into a moment that has passed.
- The composer writes a separate caption for every network in one pass. A post no longer has to fit the tightest network and read short everywhere else.
- Instagram now says up front that it publishes only from a Business or Creator account. Before, it failed quietly on Instagram's own screen.
- Settings, Account gained Export and Delete, the two controls the privacy policy promised.
- Every screen has a phone layout, with the composer and Settings as full screen sheets.
The changelog carries every release since the first public one on 22 July 2026, with a date on each.
What we are building right now
We are building generation that reads what actually performed and lets it shape what Munshi proposes next.
We are also building agent access. You hand Claude Code, Cursor or another agent a key, and it plans, drafts, schedules and reads results through Munshi. It uses the same calendar and allowance you already see. Agent access is built and running as a staff preview while the edges are worn in.
Shopify as a blog destination is built and verified against a live store. It is waiting on Shopify's own app review, which is not ours to schedule, so it is not switched on for anyone yet.
What is next, and what we will never build
Munshi now publishes directly to Facebook, Instagram, Threads and YouTube. Direct publishing to LinkedIn is built and waiting in LinkedIn's review queue. TikTok is built too, and we are preparing its submission to that review.
After that, these are the things we are considering:
- A third seat with a reviewer role.
- More blog platforms. HubSpot, Wix and Shopify blogs come up most often.
- An API and an MCP server.
- A full year of results history per channel.
None of them has a date, and we would rather say so than invent one.
Three things are settled the other way. Nothing publishes without a person's approval, and there will be no lifetime deals. Cancelling stays one click in the billing portal, with no retention flow. The roadmap holds all of it, including the never list.
The price ladder, and where it stands
The founding price climbs as seats fill, and every member keeps the price they joined at for life. Exactly one stop is open at a time. A stop is only marked filled when it has actually filled.
- Free, first 10 members (filled)
- $15 a month, first 100 members (open now)
- $29 a month, first 500 members (not yet)
- $39 a month, first 1,000 members (not yet)
- $49 a month, full price (not yet)
Today the open stop is $15 a month for the first 100 members. Studio follows the same steps at its own level. The prices, the trial and what each plan includes are on the pricing page. See the plans.
Where we got it wrong
Half the changelog is fixes, and some of them were our own mistakes. These are the ones we own.
- For three days, a check on the upload step refused anything larger than a few kilobytes. That stopped video reaching any network.
- Posts scheduled to a directly connected TikTok or Pinterest account went out the moment they were scheduled. Neither network has a scheduler of its own, and we had not built the wait.
- TikTok posts landed visible only to the poster, because the visibility chosen in the composer was not carried through.
- The picture chosen in the Featured Image step was saved and then ignored, so no article could carry a thumbnail to your site.
- A blog publish used to report success whether or not the CMS had accepted it.
We also removed a feature rather than defend it. Munshi used to propose social content of its own. The Ideas tab, the weekly batch behind it and the step that turned an idea into a draft are gone.
Social now holds Drafts, Scheduled, Posted and Replies. Writing your own post and having Munshi shape it for each network is unchanged.
What is still unfinished
Parts of the reporting we want are not built yet. Results covers the recent past in detail rather than a full year, so this September cannot be read against last September.
A workspace has two seats and no role hierarchy, so a team of three has nowhere to put the third person. Both are on the roadmap under consideration, and neither has a date.
Follow along
The changelog lists what ships. The roadmap lists what is committed and what is refused. The blog is the engine's public output, and the contact page reaches the people actually writing the code.