Agents
One set of rules for every agent you use.
A vendor can govern its own agent. Only something that belongs to none of them can govern all of them the same way. Each page says what trackline sees in that agent, and where it sees less.
Claude Code→
trackline runs as a Claude Code hook: it sees every action before it runs, and can stop one and tell the agent why.
Stops an action: yes · Tells the agent why: yes
Codex→
trackline runs as a Codex hook: it sees every action before it runs, and can stop one and tell the agent why.
Stops an action: yes · Tells the agent why: yes
Cursor→
trackline runs as a Cursor hook: it sees every action before it runs, and can stop one and tell the agent why.
Stops an action: yes · Tells the agent why: yes
Any MCP client→
trackline runs as an MCP server the agent can ask before it acts. It advises: it cannot stop an agent that does not ask.
Stops an action: no · Tells the agent why: yes
Production traces→
trackline reads the OpenTelemetry traces a deployed agent already sends. It alerts after the fact: a trace cannot be stopped.
Stops an action: no · Tells the agent why: no