8treenet/goraven

What it solves

GoRaven transforms the AI chat experience from a simple conversation window into a functional workstation for teams. It solves the problem of AI agents providing only text suggestions by enabling them to actually execute tasks—such as writing code, running shell commands, and calling APIs—within isolated, self-hosted workspaces.

How it works

GoRaven acts as an "Agent OS" that integrates LLMs with operating system capabilities and runtime environments. It allows agents to interact with files, databases, and private services. The system utilizes RAG (Retrieval-Augmented Generation) to incorporate business documents and policies into the agent's context in real-time, and supports a "Skill Marketplace" where prompts and workflows can be packaged and shared across a team.

Who it’s for

It is designed for teams who need a self-hosted, collaborative AI environment where multiple users have independent workspaces but can share project resources and skill libraries, all while maintaining administrative control over data access and model quotas.

Highlights

• Task Execution: Agents can read/write files, execute shell commands, and call MCP tool chains.
• Team Collaboration: Isolated workspaces for individuals with shared skill libraries and centralized admin controls.
• Skill Marketplace: One-click installation of packaged prompts, scripts, and workflows.
• Self-Hosted & Private: Deployable via Docker; compatible with various APIs including OpenAI, Claude, DeepSeek, Qwen, and GLM.
• Extensibility: Plugin hooks allow custom logic injection into conversations and tool calls without modifying core code.