agorOrchestrate AI Coding Sessions on a Shared Canvas

In an era where AI coding assistants are becoming ubiquitous, a critical question emerges for development teams: how do you effectively manage multiple AI sessions, maintain consistent context, and track code changes simultaneously? The open-source agor project steps in to address this challenge, proposing a 'multiplayer canvas' approach. It orchestrates AI coding sessions from models like Claude Code, Codex, and Gemini into a single, shared workspace, transforming AI-assisted development from a solo endeavor into a collaborative team effort.

A Collaborative Canvas for AI Orchestration

At its core, agor offers a real-time collaborative canvas. Think of it like Figma, but instead of design mockups, you're editing and interacting with AI coding sessions. Each team member can initiate or join an AI session on this canvas. The real power lies in what's happening behind the scenes: these sessions are powered by actual AI coding models. You could have Claude optimizing a function while Gemini generates unit tests, with all conversations and results synchronized in real-time across the canvas.

A standout feature of agor is its intelligent Git worktree management. As AI models generate code changes, the system automatically creates dedicated Git worktrees. This crucial mechanism prevents modifications from different AI agents from clashing or overwriting each other. Developers can even roll back to a specific version of an AI session, reviewing the context and generated output at that precise moment. For teams engaged in collaborative or long-running AI-assisted projects, this capability is a game-changer, mitigating what could otherwise be a significant source of headaches.

Visualizing Your 'AI Team' in Real-Time

agor meticulously records every AI interaction, presenting the complete dialogue history in a timeline or card format. This means you can always revisit why a particular AI made a certain code suggestion, or even pause one AI's task and hand it over to another for continuation. This robust conversation visualization transforms AI from a black box into a transparent, auditable, and traceable collaborative partner. It fosters trust and understanding within the team regarding AI decisions.

Under the hood, agor is built with TypeScript, featuring a React frontend and a Node.js backend. Deployment is straightforward: a simple git clone && npm install && npm run dev gets it running locally. For team usage, agor includes basic user management, currently relying on the local file system, though future integrations with external authentication systems are likely.

Practical Use Case: AI-Driven Code Review and Iteration

Imagine your team is refactoring a legacy module. With agor, you could assign Claude to analyze dependencies, task Codex with generating the refactored code, and have Gemini check for edge cases—all within the same shared canvas. If a colleague discovers a bug, they can immediately create a new AI session on the canvas, fix it on the spot, and commit the changes to a dedicated worktree. The entire process remains transparent, allowing every team member to observe and understand the AI's decision-making process and the resulting code modifications.

• Multiplayer Real-time Collaboration: All modifications on the canvas are broadcast to connected users, complete with synchronized cursor positions.
• Isolated AI Sessions: Each AI model or task operates independently, preventing interference, yet they share a common project context.
• Traceable Change History: All AI-generated code changes are linked directly to their respective conversations, simplifying review and auditing.
• Pluggable Architecture: Designed for future expansion, allowing integration of more AI backends or custom workflows.

Current Limitations and What's Next

agor is still in its nascent stages. The documentation is somewhat sparse, and certain features, such as robust user authentication and external storage options, are not yet fully developed. Developers unfamiliar with Git worktrees might face a slightly steeper learning curve. Furthermore, running multiple AI model backends locally can demand significant hardware resources. However, given its open-source nature and active development, these areas are expected to improve over time.

If you're part of a mid-to-large development team heavily utilizing AI coding assistants, or if you frequently need to compare the outputs of different AI models in parallel, agor is definitely worth exploring. It distinguishes itself from traditional IDE plugins by offering an independent, dedicated collaborative space, truly integrating AI as a core member of your development team rather than just a personal helper.