A curated list of AI tools and resources for developers, see the AI Resources .

Director

Director provides MCP-based Playbook management and runtime so agents can be declaratively configured and run locally or in containers.

Detailed Introduction

Director is an MCP-focused Playbook management and runtime platform that lets you provide tools, prompts, and configuration to AI agents via declarative YAML playbooks. Designed as local-first, Director supports one-click connections to clients like Claude, Cursor, and VSCode, and uses unified OAuth and tool filtering to minimize context and improve safety. Playbooks can be edited in the UI, CLI, or TypeScript SDK and are easy to version, share, and audit.

Main Features

  • Playbooks: portable files that define tools, prompts, and configs for agent capabilities.
  • One-click integration: connect supported clients (Claude, Cursor, VSCode) quickly.
  • Local-first and isolated: run locally or in containers with centralized JSON logging and strong isolation.
  • Multiple interfaces: Web Studio, CLI, and TypeScript SDK for interactive and automated workflows.

Use Cases

Director fits scenarios that require declarative management, auditability, sharing, and reproducibility of agent capabilities—such as multi-agent research platforms, team-level toolchain integration, and controlled automated workflows. It’s especially useful for researchers, SREs, and platform engineers.

Technical Features

The project is implemented primarily in TypeScript within a monorepo that contains multiple apps and SDKs, and offers Docker images and binary releases. Director implements the MCP spec, supports tool filtering, unified OAuth, centralized logging, and a plugin-friendly architecture to enable local and cloud deployments.

Comments

Director
Resource Info
🌱 Open Source 🧩 MCP 💻 CLI 🛠️ Dev Tools