Report Abuse

Basic Information

Common Ground is an open-source application and framework for building, observing, and collaborating with teams of AI agents to tackle complex research and analysis tasks. It provides an out-of-the-box multi-agent architecture that models a consulting team with Partner, Principal, and Associate roles, a real-time web interface for user interaction, and tooling to run, inspect, and manage multi-step agent workflows. The project supports both a recommended Docker deployment that includes a Gemini bridge and a developer local setup with Python and Node.js, enabling modification of core services, LLM providers, and frontend behavior. It emphasizes transparency and human-in-the-loop collaboration by surfacing agent decisions, tool calls, and state changes. The repository includes declarative YAML-based agent profiles, a Model Context Protocol integration point for standardized tools, built-in project and knowledge management with an updating RAG index, and is licensed under Apache-2.0.

Links

App Details

Features
The README highlights an advanced multi-agent Partner-Principal-Associate architecture for planning, decomposing, and executing work. Declarative agent design via human-readable YAML files allows no-code or low-code customization of role behavior, tool access policies, flow deciders, and handover protocols. A full observability suite offers real-time Flow, Kanban, and Timeline views plus streaming status and thoughts over WebSocket to make agent reasoning inspectable. Extensible tooling is provided through a Model Context Protocol and a Python tool system where developers add custom tools by subclassing BaseToolNode and registering with a tool registry. The framework is model-agnostic and ships with LiteLLM plus a configured Gemini bridge; LLMs and models are configurable through llm_configs. Built-in project and knowledge management supports file handling and an auto-updating RAG index. Deployment options include Docker compose and a developer local workflow for backend and frontend.
Use Cases
Common Ground helps teams and developers build robust multi-agent solutions for multi-step, research-heavy tasks by providing a structured architecture, observability, and extensibility. Users can design strategic and execution roles to split planning from execution, watch live execution flows to debug and understand decisions, and organize runs into projects with synchronized state across sessions. Developers benefit from declarative YAML agent profiles, configurable LLM providers, and clear extension points for custom Python tools and handover protocols. The integrated RAG index and file management give agents relevant context from the workspace, reducing manual prompt engineering. Deployment instructions support a Docker-based quick start with Gemini integration and a developer mode for customizing core logic and frontend behavior, making it practical to prototype, iterate, and collaborate on complex AI agent workflows.

Please fill the required fields*