Not another skill store: one vendor-neutral playbook for the whole job — skills, MCP servers, instructions, memory, and credentials your agent uses without seeing — deployed to Claude Code, Cursor, Codex, Antigravity, and Hermes in one command.
Open source and self-hostable. For software agents today, physical robots tomorrow. 🦉
A Playbook is the complete operating environment for an AI agent - everything it needs to work effectively, all in one place.
Think of it as your agent's portable work profile: a source of truth that travels with you across platforms and vendors.
Personality & Style
The agent's base personality, communication style, and system prompt. This defines HOW the agent behaves and responds.
Knowledge & Capabilities
Structured task definitions and knowledge the agent can use. Skills define WHAT the agent knows and can do — and can bundle files, or point at recordings of the task being performed, for agents that learn by watching.
Tools & Integrations
External tools and APIs the agent can access via Model Context Protocol. Code execution happens at the agent itself - these are the connections.
Working Documents
Structured markdown documents that serve as the agent's workspace. Notes, drafts, project files - persistent documents the agent can read and write.
Persistent Context
Key-value storage for remembering facts, preferences, and state across sessions. The agent's long-term memory.
Define your personas, skills, and MCP configurations in one place
Get a unique URL that serves your playbook in any format
Give the URL to your GPT, Claude, or any AI agent
Real problems that AgentPlaybooks solves
Want to switch from ChatGPT to Gemini, but can't take your memory and conversations with you?
Your playbook travels with you. All personas, skills, and memories are stored platform-independently - just point your new AI to your playbook URL.
Built amazing skills in Claude, but want to use them in other platforms too?
Skills in your playbook work everywhere. Export in Anthropic format, OpenAPI for GPTs, or any other format - your investment isn't locked to one vendor.
Love Grok but miss ChatGPT's custom actions and integrations?
Define your MCP servers and tools in your playbook - any AI that supports tool use can access them, regardless of platform.
Your team uses different AI platforms, but needs consistent behavior?
Share a playbook URL. Everyone's AI - whether Claude, GPT, or local LLM - follows the same guidelines and has access to the same skills.
Worried about being locked into today's AI platform?
AgentPlaybooks is open source with standard formats, and already ready for agents with a body: a skill can carry demonstrations — timestamped recordings of the task, in the order a robot performs them.
Running local LLMs but want the same capabilities as cloud services?
Point your local model to your playbook. Same skills, same persona, same memory - just running on your hardware.
Zero vendor lock-in. Portable skills. Self-hostable. The knowledge your agents accumulate is valuable - keep it yours.
Browse, share, and discover community-created skills. Like an app store, but for AI capabilities.
Run on your infrastructure - on-premise, private cloud, or air-gapped. Full control, full privacy.
Works with Claude, ChatGPT, local LLMs, and future robots. Switch platforms without losing knowledge.
Browse public skills, personas, and MCP servers built by the community, and install any of them into your own playbook. For software agents today, physical robots tomorrow.
AgentPlaybooks is a portable operating environment for AI agents. A playbook holds everything an agent needs to work — its persona, its skills, its MCP servers, its project instructions, and its memory — as one source of truth that belongs to you rather than to any single vendor.
Across supported coding agents, the same portable configuration can be rendered for Claude Code, Cursor, Codex, GitHub Copilot, Gemini CLI, and others. Drift is reported before copies diverge. Backups carry secret references, not credential values; those stay in your environment or the encrypted vault.
Run apb doctor to audit, apb push to save a private, versioned portable snapshot, and apb pull plus apb sync to restore Agent Skills, custom agents, MCP references, and project instructions on another supported platform. Complete skill folders, including scripts and assets, are preserved.Read the backup and compatibility guide.
Open source and self-hostable, so the playbook is the portable source of truth — not any single vendor or editor. Switch platforms, run several at once, or host the whole thing yourself. Built for coding agents and AI assistants today, and for the physical robots that will need the same portable skills tomorrow.