Integrations
Passport
- Rock
- Locks
- Depth
- 2 · Excavator
- Time to dig
- 1–3 years
- Capital
- ◐ · medium
- Solo
- ✓ solo-reachable
- AI
- ↓ under attack
- Rent
- ~ partly
Sample
- Share of apps
- 40.2%
- No-rate
- 29%
- Median price
- $19.99
Figures from the canivibecodeit sample. No-rate is the share of apps carrying this tag that cannot be vibe-coded — a proxy for structural strength. Only the first thirteen mechanics were measured.
Essence
The breadth of your connectors and the endless work of keeping them alive: OAuth, calendars, banks, plugin ecosystems, platform partnerships. A boring moat, but historically an effective one — nobody wants to rebuild a thousand connectors and, more to the point, to repair them forever.
How it is built
Zapier — the long tail as the product
Seven thousand integrations, most of them barely used, and yet it is the tail that creates the position of "everything is here". Half the moat is not the writing but the stubborn operation of it: APIs change, tokens expire, webhooks break. Plus an ecosystem trick — app developers build the integration with Zapier themselves to get into its catalogue, so the moat is dug by other people's hands.
Plaid — fragile integrations as the barrier
Thousands of bank connections, many of them historically screen-scraped and broken by every redesign of an online bank. The barrier is not the difficulty of one integration but the operational machine that keeps all of them running at once, plus the contractual relationships with the banks.
Segment and Calendly — depth instead of breadth
Calendly is not "a connection to a calendar" but years of handled edge cases: time zones, reschedules, group events, conflicts. The depth of a single integration, taken all the way to reliability, is also a moat — and one that is harder to see from outside than the size of a catalogue.
How it is bypassed
Standardisation — PSD2 and MCP
When a standard appears, a moat built of proprietary connections melts: Europe's PSD2 obliged banks to expose APIs, Plaid's scraping advantage in the EU lost its value, and competitors grew on the open interfaces. MCP is starting the same process in tool integrations: once applications publish a standard interface themselves, a catalogue of 7,000 connectors turns from a moat into a museum.
AI writes and repairs the connectors
What cost Zapier thousands of person-hours is now generated by agents: writing a client from API documentation is a task an LLM does in minutes, and repairing a broken connector can be automated from the logs. Computer-use agents go further and work through the UI where there is no API at all. The cost of reproducing the moat falls by orders of magnitude.
The head of demand — Make against Zapier
Twenty integrations cover 80% of real usage. The attacker builds only those and wins on price and quality exactly where the demand lives: Make pressed Zapier on price in mass scenarios, and platforms finish the job from below by building the popular integrations in natively. The incumbent keeps its tail — along with the cost of maintaining it.
Verdict
The moat is dug with the tail, with operations and with the depth of edge cases. It is bypassed by standards, by AI-generated connectors and by an attack on the head of demand — in the AI era, one of the fastest-cheapening moats in the atlas.