Oliver Eidel · February 23, 2026

Your SaaS Will Have an API, Whether You Build It Or Not

Agentic Coding
With agents on the rise, there will be big changes in how people interact with SaaS in the future. So what will this mean for SaaS companies?
  • Agents are here to stay: Interacting with Claude Code / Codex is way more efficient than clicking through a clunky SaaS UI yourself.
  • If you don't build an API at all, agents will simply use your UI: Even if you retreat to some weird corporate position of "we want to own the user experience and we won't build a REST / MCP API", the problem you're overlooking is that your product already has an API, which is your current human-facing UI. As Peter Steinberger correctly notes, agents will simply start interacting with your UI through the browser, which will effectively mean that you simply have implemented an API which is somewhat mediocre, slow and rate-limited, because browser <--> agent interactions are still somewhat slow (for now).
  • If you don't build a good API, your competitors will: In the future, SaaS purchasing decisions will be driven by how compatible a product is with the agents a person / company is using. If your competitor offers a product which is agent-first with a good API, then people will prefer choosing your competitor.
  • Some SaaS products will be "downgraded" to API backends: With mostly agents interacting with products, there's only limited need for a human-facing UI. Accordingly, many products might end up becoming agent backends. The big question is whether they will survive at all, see next point.
  • Some SaaS products will disappear entirely: With most SaaS becoming API backends, the discussion will shift towards whether certain products are required at all. The big question, as Steve Yegge notes, will become whether it's cheaper to purchase a SaaS product (xx€ / month) vs. building it yourself (xx€ in tokens + your time + (xx€ in hosting + maintenace + tokens / months)).
    Let's take the Gmail part of Google Workspace as an example: Sure, you could build your own email software, but maintaining an email server and getting it whitelisted by spamfilters is notoriously hard, and Google Workspace is not that expensive (at least per user), so purchasing it makes sense.
    But this doesn't apply to all SaaS products - I see a lot of crappy "enterprise software" with opaque pricing ("contact sales") and bloated teams (yep - "contact sales") going away.

.. and then there'll be more second-order consequences which are hard to predict right now. It'll be interesting to which SaaS products exactly are going to go away, what this will mean for pricing of the surviving products, and how much of a role the "human element" will play in the future: Humans might prefer to purchase products where they can reach out to a human to receive help if they get stuck. But then again, none of this might be rational, because agents would do a better job. But then again, a lot of SaaS purchasing nowadays is highly irrational, especially in the enterprise market, where mostly-useless (human) demo calls and customized pricing seems to be the standard. What will happen to that?

We'll see. What a time to be alive.

Leave a comment. Or send me a message on LinkedIn.