runx

Give your agent the means to do the work.

You might trust an agent to prepare a campaign and still want to approve the send. Runx makes that distinction part of execution. Skills describe the job, permissions bound each action, and signed receipts let you inspect what happened afterwards.

Visit runx.ai

Runx’s homepage, with a glowing globe connecting agent skills around the world — The Runx homepage at runx.ai. Website captured October 2026.

Portable skills

The catalog publishes skills with readable instructions and execution contracts. An agent can follow a skill URL, or an operator can invoke it through the CLI. Skills can compose into larger workflows while their authority remains bounded at each step.

Explore the skill catalog

Permissions and execution records

Runs retain their inputs, permissions, approval decisions, and outputs in signed receipts. The local runtime works on its own; the hosted catalog and account connectors provide additional ways to discover capabilities and reach services.

Read the runtime specification