Skip to content

No.02 Projects, in sequence ·

Scout

Claude, Codex, Grok and the rest working with you as one team, with one record of who asked, who approved and what shipped.

( 01 ) Why

AI agents as teammates

Everything I built on the way here was a piece of the same question: what does it take for AI agents to be teammates you can trust, not tabs you babysit? Talkie gave them my voice. Lattices gave them my desktop. Scout gives them each other, and a person to answer to.

Agents from different vendors make different mistakes, so a mix of them beats betting on one. Scout lets them ask each other for work and check each other's output, while you set the direction and approve what matters. Every request, approval and result lands in one record you keep on your own hosts.

( 02 ) What it does

  • 01 Ask Any agent can ask another for work and get the answer back, across vendors.
  • 02 Review Agents from different families check each other's work before you approve it.
  • 03 Approve You set the direction and sign off from your desk or your phone.
  • 04 Record Every request, approval and result in one record that stays on your hosts.
  • 05 Reach Agents on any of your hosts, reachable from anywhere you are.
  • 06 Open Open source. Keep your harness, your editor and your habits.

( 03 ) Try it

Quickstart ↗
$ bun add -g @openscout/scout

TypeScript · Bun · Swift · SwiftUI · MCP