5 questions that tell you which of your tools become AI infrastructure + 3 prompts to run the audit

Source: Nate’s Newsletter Date: 2026-05-02 URL: https://natesnewsletter.substack.com/p/issue-trackers-agent-infrastructure

Summary

Nate’s framework distinguishes enterprise tools that will become agent infrastructure from those that get “wrapped” or replaced. The signal is Linear: its CEO declared issue tracking dead in March 2026, yet when OpenAI open-sourced Symphony (an autonomous coding system), Linear became its literal control plane — with some teams seeing 500% increases in pull requests. The durable parts aren’t the UI; they’re the state machine, assignee field, audit history, and dependency graph — exactly what agents need to coordinate work.

Implications

• Agent infrastructure taxonomy. The five-questions framework gives teams a lens for evaluating their existing tool stack before agents arrive — which tools survive as coordination substrate versus which get wrapped into model context. This is the practical operationalization of the broader agent-native software thesis.
• Issue trackers as control planes. Linear’s accidental architecture (structured state, typed fields, persistent history) made it a near-optimal substrate for autonomous coding agents. The pattern extends to CRMs, ERPs, and service desks: tools built for human coordination often have the data shapes agents need.
• MCP as the integration primitive. The article’s practical output (three diagnostic prompts + MCP server specs) positions Model Context Protocol as the standard interface through which agent-ready tools expose themselves. MCP server count becomes a proxy for “agent-ready” status in tool selection.