Even when you do everything right inside your own mesh, the components you depend on — external APIs, SaaS tools, vendor systems — are largely not ready for this world. Most commercial software was built for humans clicking through UIs, not for agents making programmatic decisions at speed.
Connect your auth system to your CRM so signups become leads, and you discover the CRM API has no sorting, no date filtering, no way to retrieve recent records without pulling the entire dataset. There are no event hooks. The pagination is inconsistent. Authentication assumes a user session, not a long-running service. This is the state of most commercial APIs when you try to use them as components in an automated system.
The vendors will eventually fix this. The pressure is building. Customers are wiring up agents to every product they own and discovering which vendors built real APIs and which ones built marketing glossaries. The next decade of SaaS competition is going to be settled on whether the API can serve an agent at machine speed, not whether the UI is pretty.
In the meantime, the agent mesh is also a pattern for insulating your system from the fragility of external dependencies. Each atomic agent that wraps an external service is a replaceable unit. When the vendor finally fixes their API, or you switch vendors, you swap one agent. The rest of the mesh does not care. This is the same property I described in the declarative atomic agent — small composable units make the system robust to changes you cannot control.
If you are integrating against vendor APIs, treat each integration as a soft dependency. Wrap it. Stub it for tests. Be ready to replace it. The vendors that survive the agent transition will be the ones who fix their APIs in the next eighteen months. The rest will lose their seat at the table because their customers' agents started routing around them.
Sources
Related Essays
The Atomic Agent Mesh: Architecture, Build-vs-Buy, and the Review Layer
Enterprise AI will not be one mega-agent. It will be a mesh of atomic, auditable units, and the companies that nail review and context will own the next infrastructure layer.
Triggered Workflows Generate Most of the Volume
Most enterprise agent value comes from background workflows, not from humans typing into a chat box. Machines do not sleep. Lean into triggered work or get out-shipped.
The Orchestration Bet
Most companies are betting on building the best agent. A smaller number are betting that the best agent changes every week — and that the durable value is in orchestration.
Key takeaways
- Most commercial APIs were built for human-driven UIs, not agent-driven automation.
- Missing sorting, filtering, pagination, and event hooks turn straightforward integrations into hacks.
- Wrap each external dependency in an atomic agent so you can swap vendors without touching the rest of the mesh.
FAQ
Why are APIs not ready for agents?
Because the API was the second-class citizen behind the UI. Most vendors shipped read endpoints with no filtering, no sorting, no efficient way to retrieve recent records. That is fine for humans. It is broken for agents.
How does the mesh insulate you from this?
Each atomic agent that wraps an external service is a replaceable unit. When the vendor finally fixes their API, or you switch vendors, you swap one agent. The rest of the mesh does not care.