Most data platforms today operate in isolation.
Coordinating changes across them is still largely manual.For example, if a schema changes in one system, engineers often need to manually update pipelines, dashboards, and downstream consumers.
We started exploring a different idea.
Instead of building more integration layers, what if each data platform had an autonomous agent that understood it deeply?The idea we're experimenting with is something like an Agentic Operating System for data platforms.
Each platform would have an agent responsible for understanding:
schemas
dependencies
pipelines
system changes
historical operations
The agents themselves are open sourced so organizations can run and modify them without vendor lock-in.The interesting part is what we call the agentic brain.
Most AI assistants rely on conversation history for context. That works for chat, but infrastructure systems require long-term understanding.
So instead of relying on chat context, these agents retain persistent memories about system state and historical changes.
Because of this, an agent can reason about things like:
how a system evolved
what downstream systems depend on
what changes might break other platforms
In theory, if a schema changes in one system, the agent could:
detect the change
analyze downstream dependencies
coordinate with other platform agents
plan the required updates.The long-term goal is infrastructure where platforms coordinate automatically instead of relying on manual integration work.
We filed a patent for the architecture last year (currently pending), but the agents themselves will remain open source.
Curious if others here think agent-based coordination across infrastructure could work in practice.