Ticketing Systems Are Infrastructure for Coding Agents

Ticketing systems are becoming infrastructure for coding agents, not tools for teams. Linear knows it. OpenAI knows it. And if the rumors about Anthropic acquiring Atlassian for $40 billion have any substance, the whole industry is starting to price it in.

The shift is subtle but total. Agents don't need issue tracking for visibility or team coordination. They need it as operating memory. State handoffs, execution context, continuous task loops, all of it breaks down when the ticketing layer can't keep pace with how agents actually work.

I built a framework to run coding agents continuously for over twenty hours. When I tried GitHub Issues as the state layer, it fell apart fast. Rate limiting under agent speed, schema rigidity that required upgrading to GraphQL just to support custom fields, friction everywhere. I reverted to the file system, which worked because it's immediate and has zero overhead. But that came with its own problem: no visibility. No way to see the overall project state at a glance, no logical relationships between tasks, no reporting, no easy navigation across a complex multi-part project. It works for the agent. It doesn't work for the human trying to stay in the loop.

That tension is exactly what makes this infrastructure problem hard. The ticketing system needs to serve two masters simultaneously: fast enough and structured enough for agents to use as operating memory, while also giving humans the visual clarity and navigational context to understand what's actually happening across the whole system.

Linear's "issue tracking is dead" announcement is essentially making the same argument. The next system isn't built around human handoffs, it's built around agent context. OpenAI's Symphony framework uses Linear as its state backbone, and that's not a coincidence.

The teams that figure out this layer early are going to run circles around everyone still treating ticketing as an afterthought.