@brandonducar5
Brandon DucarCreator▲9
Founder/Ceo, Ducar Consulting | DreamNet
32m ago · New project · Agents · Palm Beach

Shipped: DreamLoops

Your agent shouldn’t have to start from scratch every session—or become a different agent just to learn something new.

DreamLoops combines persistent operating loops with Capsules: portable packages of capabilities, knowledge references, policies and evaluation criteria that can be added to an existing agent.

Four connected pieces make it work:

  • DreamLoops define repeatable work: triggers, context, permitted handlers, limits and completion checks.
  • Capsules package capabilities that can be validated, versioned, composed and evaluated before activation.
  • Receipts connect recorded execution to its actor, checks, approvals and results.
  • Lineage tracks where capabilities came from, how they changed and what to roll back or quarantine.

The repository includes a dependency-free runtime, CLI, persistent-agent starter kit, reusable loops and optional adapters. The starter covers restart recovery, planning, research, relationship continuity, memory consolidation, approval requests and checkpoints.

It works independently of DreamNet. You can adopt individual components rather than replace your existing agent stack.

External implementation evidence includes merged DreamLoops persistent-agent support and a Bonfire memory adapter in ZOL, an independently operated agent.

Capsules carry capability—not automatic authority. Loading one does not grant permission to execute code, publish, spend money or change production. Applications explicitly register handlers and permissions.

Explore the source, initialize a starter agent, or bring an existing agent you want to make more persistent and inspectable.

DreamLoops@brandonducar/dreamloops-2 · github.com/BrandonDucar/dreamloops
ai-agentsagent-memoryagent-orchestrationprovenanceopen-source

Comments

No comments yet
Start the conversation.
Log in to join the thread.