AgentAppFlow

Local-first infrastructure for project-aware AI development workflows.

What Is AgentAppFlow?

AgentAppFlow is an open-source system for turning AI coding tools into project-aware, guardrailed collaborators.

The product direction is simple:

• a Mac-first app handles onboarding, controls, approvals, and project visibility
• a Python orchestration core manages AI workflows, memory, and retrospectives
• a Rust execution layer enforces guarded local actions and policy boundaries
• every managed project keeps its framework state inside .agentappflow/ so the setup stays visible, portable, and local

The result is a local development workflow where AI agents can operate with project memory, repeatable rules, and controlled self-improvement without depending on a backend.

Why This Exists

Most AI coding tools are powerful but stateless, inconsistent, and weak at preserving project-specific working patterns over time.

AgentAppFlow is designed to close that gap by combining:

• local project memory inspired by tools like Claude-Mem
• specialized internal agent roles inspired by systems like HyperAgent
• explicit repo-level framework state that can evolve with the project
• approval-aware guardrails so automation stays useful without going rogue

Architecture At A Glance

Swift macOS app
  -> onboarding, project registration, approvals, history, framework health

Python core
  -> planning, memory, retrieval, retrospectives, framework proposals

Rust executor
  -> guarded execution, filesystem policy, diff validation, audit trail

Managed repo
  -> .agentappflow/

Core principles:

• local-first, no backend required
• Mac-first UX today, OS-neutral product direction long term
• Python owns AI cognition
• Rust owns risky local execution boundaries
• project memory stays visible instead of disappearing into a black box

Repository Map

This repository is the ecosystem index. The main product repo is agentappflow-app.

• agentappflow-app: primary product repo and Mac-first local control center
• agentappflow-docs: deeper architecture, memory design, and adapter conventions
• agentappflow-cli: command-line automation and scaffolding tools
• agentappflow-plugins: plugin architecture and integrations
• agentappflow-skills: reusable developer skills and workflows

Project Contract

Every managed user repo is expected to contain:

.agentappflow/
  project.yaml
  rules/
  templates/
  sessions/
  retros/
  proposals/
  cache/

This is the shared contract that lets tools like Codex and Claude Code discover project context, follow rules, reuse memory, and participate in controlled framework evolution.

Current Status

AgentAppFlow is in the architecture and product-foundation phase.

Current focus:

• defining the app-first product direction
• locking the Swift/Python/Rust runtime split
• designing the .agentappflow/ project framework layout
• preparing the first implementation slice for onboarding and project bootstrap

Branching

• main: stable baseline
• dev: active development

License

MIT. See LICENSE.

Star History