Platform

The components we are building.

LOOP Automation Engine (AFEE)

Bounded work, carried from plan to accepted outcome.

Each workstream gets an isolated lane with an AI conductor. Software owns the workflow and its gates; models advise and execute bounded steps, and every verdict is bound to an exact commit.

In DevelopmentInternal use

  • Isolated lanes: one writer per worktree, with explicit budgets and leases.
  • A conductor that sequences analysis, implementation, tests and independent review.
  • Lane types for software builds, code review with bug fixing, and brainstorming.

Memory Fabric (AFMF)

Agent memory that is reviewed before it is reused.

Run records, decisions and lessons become shared agent memory. Agents recall it over MCP, so any compatible agent draws on the same reviewed history instead of a growing transcript.

In Development

  • Hybrid recall: keyword and semantic search, combined into one ranking.
  • Served over the Model Context Protocol (MCP).
  • Review before reuse, so one mistake does not spread to every lane.

Foundry

The factory for AI organizations.

Create and evolve a parallel AI organization from an approved business mapping.

It brings the other components together: lanes do the work, memory carries the lessons, and people keep the authority.

Vision

  • Roles and authority mapped from your own organization.
  • AI teams built level by level from that mapping.
  • Human approval at every level of authority.