<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> This project addresses the challenge of maintaining AI agent knowledge and skills across diverse development environments. It provides a portable .agent/ folder containing memory, skills, and protocols that can be plugged into various tools like Claude Code, Cursor, and others, enabling agents to retain their learned information and configurations regardless of the underlying platform. This benefits developers and researchers by ensuring consistent AI agent behavior and knowledge persistence across different toolchains.

How It Works

The core innovation is the .agent/ folder, acting as a self-contained "brain" for an AI agent. This folder includes distinct memory layers (working, episodic, semantic, personal) with query-aware retrieval and nightly compression into candidate lessons. A crucial component is the host-agent review protocol, where a CLI-driven process (graduate.py, reject.py) ensures learned lessons are explicitly reviewed and rationalized before being committed to semantic memory, preventing unattended reasoning and provider coupling. Skills are loaded progressively based on task triggers, and permissions are enforced via a permissions.md file.

Quick Start & Requirements

• macOS/Linux: Install via Homebrew: brew tap codejunkie99/agentic-stack then brew install agentic-stack. Navigate to your project directory and run agentic-stack claude-code (or other adapters like cursor, windsurf, opencode, openclaw, hermes, pi, standalone-python).
• Windows (PowerShell): Clone the repository (git clone https://github.com/codejunkie99/agentic-stack.git), navigate into it (cd agentic-stack), and run the native installer: .\install.ps1 claude-code C:\path\to\your-project.
• Prerequisites: Homebrew is required for macOS/Linux installation. Python is implicitly needed for the standalone adapter and scripts.
• Onboarding: An automatic terminal wizard guides users through setting preferences (e.g., agent name, language, explanation style, test strategy, commit style, code review depth) and optionally enabling features like FTS memory search.
• Links: The project's GitHub repository serves as the primary source for documentation and installation.

Highlighted Details

• Harness-Agnosticism: A single .agent/ folder (memory, skills, protocols) functions across multiple AI coding assistants and DIY Python loops.
• Portable AI Brain: The .agent/ folder encapsulates the agent's state, enabling knowledge persistence across different environments.
• Host-Agent Review Protocol: A mandatory CLI-driven review process (graduate.py, reject.py) for learned lessons ensures explicit human oversight and rationale.
• [BETA] FTS Memory Search: An opt-in full-text search capability over all memory documents using FTS5, falling back to ripgrep or grep.

Maintenance & Community

The project is actively maintained, with recent updates noted (v0.6.0, v0.5.0). The author is @AV1DLIVE. Specific community channels like Discord or Slack are not detailed in the provided README.

Licensing & Compatibility

• License: MIT License.
• Compatibility: The MIT license generally permits commercial use and linking within closed-source projects without significant restrictions.

Limitations & Caveats

The FTS memory search feature is explicitly marked as [BETA]. Users migrating from the older OpenClient adapter to OpenClaw must re-run the installer (./install.sh openclaw). The detailed memory layers and review protocol suggest a potentially complex setup or debugging process for advanced users.