Description
agentmash is a desktop application that unifies and coordinates various AI coding agents for development teams. It acts as a central hub, allowing agents such as Claude Code, Codex, Cursor, GitHub Copilot, and others to operate on their respective logins while being aware of each other's activities within a shared repository. This coordination is crucial for preventing merge conflicts and enhancing collaborative coding workflows.
Read More
The core functionality of agentmash revolves around real-time communication and conflict detection among AI agents. Before an agent edits a file, agentmash informs it if a teammate's agent is already working on that same file or branch. This proactive warning system, referred to as "antigravity," helps agents avoid overwriting changes and reduces the need for manual conflict resolution after merges. The platform also allows agents to "claim" files, providing an advisory notice to other agents about intended work areas.
agentmash supports a wide array of coding agents, running each in its own terminal or chat interface. It integrates with popular tools like Claude Code, Codex, Gemini CLI, Cursor, Copilot, OpenCode, Goose, and Antigravity, and can assist in installing those not yet present. The application ensures security by never accessing agents' API keys, maintaining each developer's control over their individual agent subscriptions.
For team collaboration, agentmash offers a live view of who is online and what each agent is doing, displayed on the side of the application window. Developers can click on a teammate to view their weekly activity, including sessions, files, commits, and claims. Teammates can be added using their GitHub names. The platform also enables agents to ask questions of a teammate's agent, with prompts requiring approval on the teammate's machine before execution, ensuring controlled interaction.
Privacy is a key consideration; agentmash ensures that file contents and code remain on the developer's machine unless a session is explicitly shared live, in which case data is encrypted end-to-end. Only metadata like file paths, branch names, line counts, and the first 500 characters of prompts are shared with the room. For teams requiring self-hosting, an Enterprise plan offers a private Docker image, ensuring prompt excerpts never leave the network.
agentmash offers flexible pricing, including a free tier for up to two developers, a Team plan billed monthly or annually per developer, and an Enterprise plan for larger organizations with custom requirements like self-hosting, SSO, and audit export. The desktop app, agentmash lobby, is included with every plan, emphasizing that users pay for developers, not for the app or agents themselves.
- CategoryAI Coding Assistants
- Pricing ModelFreemium
- DeploymentDesktop App, Self-hosted (Enterprise)
- PlatformsmacOS, Windows
agentmash's Core Features
Real-time AI agent coordination
Cross-agent conflict detection and warnings
Support for multiple AI coding agents (Claude Code, Codex, Cursor, GitHub Copilot, Devin, etc.)
Desktop application for macOS and Windows
Live team activity monitoring and session tracking
Agent-to-agent communication with prompt approval
File claiming mechanism to signal work areas
Privacy-focused: code stays on local machine by default
Integration with Git for commit reporting
Pre-edit advisories and post-edit reporting
Support for public and private repositories
Ready-made Claude Code agent teams (Crew)
Self-hosting option for Enterprise plans
Unlimited agents per developer on paid plans
How to use agentmash?
Connect Repository: A user with push access connects the repository at agentmash.dev/install.
Install App or CLI: Each developer downloads the agentmash Lobby desktop app or runs `npx agentmash login` in their terminal.
Sign In: Developers sign in with GitHub to link their account.
Restart Agents: Restart coding agents to enable agentmash hooks and reporting.
Approve Hooks: For agents like Codex, approve agentmash's hooks in their respective settings.
Collaborate: Agents will now warn each other of changes and report activity in real-time.
agentmash's Use Cases
- Collaborative Coding
- Real-time Conflict Prevention
- Team Activity Monitoring
- Agent-to-Agent Interaction
- Codebase Coordination
- Streamlined Development Workflows
- Secure AI Agent Operations
- Custom AI Agent Teams
