AgentPack is an open-source orchestration harness and configuration management framework developed by Vishal (vishal2612200) for multi-agent coding environments. It standardizes agent behavior, steering rules, and tool skills across a wide range of terminal- and IDE-based AI coding agents, eliminating configuration drift across developer setups. In daily operation, AgentPack synchronizes centralized rules and capabilities to agent-specific formats, supporting integrations such as Claude Code (.claude), Cline (.clinerules), Cursor (.cursor/rules), Windsurf (.windsurf/rules), Codex, Kiro, and OpenCode. Developers define repository-level rules and modular skills once, and AgentPack distributes them to the appropriate local agent configuration directories or IDE plugin paths. Unlike standalone agent runners that manage their own execution loop in isolation, AgentPack functions as a meta-harness. It bundles multi-agent rules, benchmark suites to evaluate agent behavior across toolchains, modular skill definitions, and a local frontend dashboard to monitor agent integrations across repositories. The project is open-source on GitHub, packaged for distribution via npm and Python/source builds, and actively maintained with modular agent rule definitions.
Cross-agent rule synchronization — Translates and maintains steering rules across Cursor, Windsurf, Cline, Claude Code, OpenCode, and Codex. - Modular skill packaging — Provides a structured skill library to export consistent capabilities and tool routines across diverse coding agents. - Integrated benchmark harness — Contains a dedicated benchmark directory to measure rule adherence and task completion across different agent platforms. - Local frontend dashboard — Includes a frontend dashboard application for inspecting rules, skill sets, and agent configurations visually. - Multi-IDE and plugin adapters — Delivers native adapters for VS Code, Codex plugins, Kiro steering, and standalone terminal agents.
Use cases
Use Case: Centralizing Repository Rules Across Multi-Tool Teams Engineering teams with members using different AI coding tools (such as Cursor, Windsurf, and Claude Code) can maintain a single source of truth for architectural constraints and coding guidelines, automatically syncing them into repository dotfiles.
Use Case: Packaging and Distributing Custom Skills Teams can author domain-specific development routines—such as schema migrations, test generation patterns, or deployment checks—into portable skill bundles deployed uniformly across all developer workstations.
Use Case: Benchmarking Agent Rule Adherence Tool evaluators and lead engineers can run standardized benchmark scenarios across supported agent runtimes to determine which model-and-harness combination best follows project-specific constraints.
Who it is for
AgentPack is designed for engineering leads, developer tooling teams, and developers who work across multiple AI coding environments (Cursor, Windsurf, Cline, Claude Code) and require consistent steering rules and skills across the repository. It is less suited for developers using a single fixed IDE assistant who do not need cross-agent configuration synchronization.
Install guide
Installation Install AgentPack globally via npm or clone the repository directly for local development: bash # Install via npm (standard package installation) npm install -g agentpack # Or clone from source git clone https://github.com/vishal2612200/agentpack.git cd agentpack npm install
First Run Initialize AgentPack in your project repository to detect active agent configurations: bash # Initialize AgentPack in your current repository agentpack init # View configured rules and target agents (Cursor, Claude, Cline, Windsurf) agentpack status
Example Session Sync central agent rules to all supported agent configuration formats and run the rule validator: bash # Synchronize shared rules across .cursor/rules, .clinerules, and .windsurf/rules agentpack sync # Launch the local dashboard to inspect active skills and integrations agentpack dashboard
LoopTroop — LoopTroop is an open-source AI coding agent harness and orchestration framework maintained by the looptroop-ai organization on GitHub. Built with…
Crystal — Crystal is an open-source AI coding harness and desktop-oriented agent platform maintained by Stravu. Designed as an Electron-based application with…
SWE-agent — SWE-agent is an open-source autonomous software engineering agent harness built and maintained by researchers from Princeton University and Stanford University.…
Continue — Continue is an open-source AI code assistant and agent framework originally developed by the Continue team (continuedev) and recently acquired…
Vibe Kanban — Vibe Kanban, originally developed by Bloop AI, is an open-source orchestration harness and visual task board designed to manage and…
Hivelore — Hivelore is an open-source agent harness and contextual memory orchestrator created by Doucs91. It provides persistent project memory, decision tracking,…