AI agents speed up development — data access slows them down
The New Stack Doug Sillars ● Covered by 2 sources
AI agents are speeding up app builds. Getting them live data is the part that still bogs teams down.
Based on reporting by The New Stack, Doug Sillars — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
AI agents are making development feel a lot less like slogging through scaffolding. Teams can now finish apps and internal tools faster than before. The bottleneck has moved somewhere uglier: access to live operational data.
That access is still handled the old way. Teams open tickets with each system owner, wait for approvals, defend why they need it, and chase down unanswered emails. If the data lives in APIs, databases, or cloud services, every connection tends to be bespoke, with its own rules and its own subset of data.
And once the tool exists, the work doesn’t stop. A PostgresDB migration can turn into a scavenger hunt for every connected app. Someone leaving the company raises the same problem in reverse: has every tool been cut off? Add AI agents into production and the anxiety gets sharper, because now teams also need to know who connected them and for what purpose.
The pitch for a unified API layer is simple enough. Put one governed access layer in front of the data sources, centralize access control, and give every human, app, and agent a scoped identity. That lets different consumers ask for the same thing and still get different responses, based on what they’re allowed to see.
Monospace is one example of that model. It connects to databases like PostgreSQL, Supabase, MySQL, and MariaDB, plus SaaS platforms and internal APIs, without migrating the underlying system. Developers get one typed SDK. Agents get an MCP endpoint. The real promise isn’t that access disappears; it’s that the mess is finally pushed into one place instead of spread across every system in the company.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI that actually matters: not smarter demos, but fewer excuses for bad plumbing. Let agents loose on production data without real identity and field-level rules, and the surprise will be an incident report, not a breakthrough. The industry loves to talk about model quality; most teams still lose to permissions and tickets like it’s 2009.
Read more about this at: The New Stack