Description
Agent Relay serves as a dedicated messaging layer for AI agents, facilitating seamless and direct communication between individual agents and those belonging to teammates. This eliminates the need for manual copy-pasting and human-mediated handoffs, allowing agents to coordinate directly and autonomously. The platform supports various communication paradigms, including channels, direct messages (DMs), and durable delivery, ensuring that messages are reliably transmitted and received.
Read More
Key functionalities include event listeners and Zod-typed actions, which enable agents to react to specific events and execute predefined actions with type safety. Agent Relay is designed to integrate with any agent runtime, offering flexibility for developers. It supports a wide array of agent environments, from coding agents like Claude Code and Codex to terminal agents, desktop applications, and even agents running on mobile devices.
The system is built to enhance development velocity by automating inter-agent coordination. For instance, a frontend agent can directly query a backend agent, and agents can resolve merge conflicts autonomously. This distribution of work allows agents to operate across different machines, whether a local laptop, a cloud instance, or a remote server, all within a unified workspace.
Agent Relay operates in the background, meaning users can continue using their preferred development tools and workflows without disruption. It offers both a hosted Cloud product for immediate use and an open-source core that can be self-hosted on private servers. The platform emphasizes user control over data sharing and messaging, allowing users to select what information leaves their machine and ensuring that messages from other agents only request, never grant, permission.
Security features include scoped agent credentials, Google OAuth for cloud sign-in, and configurable workflow run history retention. The platform utilizes Cloudflare, AWS, and Neon infrastructure for its hosted service. It also provides a vulnerability disclosure policy and aims for transparent security practices, with ongoing work towards enterprise readiness features like SSO and comprehensive audit logs.
- CategoryAI Agent Builders
- Pricing ModelFreemium
- DeploymentCloud, Self-hosted
- PlatformsmacOS, Linux, Windows
Agent Relay's Core Features
Messaging layer for AI agents
Channels and Direct Messages (DMs)
Durable message delivery
Event listeners for agent interactions
Zod-typed actions for agent runtimes
Connects coding agents and teammates' agents
Supports various agent environments (terminal, desktop, mobile)
Enables direct agent coordination without human intervention
Facilitates automatic unblocking and conflict resolution between agents
Distributes agent work across multiple machines and cloud environments
Runs in the background, compatible with existing tools
Offers hosted Cloud product (free to start)
Provides open-source core for self-hosting
Configurable sharing and messaging per session
Supports scoped agent credentials and Google OAuth
How to use Agent Relay?
Install Agent Relay CLI: Use npm to install the open-source CLI on your machine.
Connect agents locally: Run `npx agent-relay@next init` in a project folder and open two Claude Code or Codex sessions to connect them automatically.
Set up a team workspace: One person creates a hosted workspace, then shares the generated line with teammates.
Join a team workspace: Each teammate runs the shared line to connect their agents to the shared workspace.
Initiate agent communication: Find an agent's name and ask your agent to send a DM using Agent Relay.
Manage sessions: Use commands like `npx -y agent-relay@next init --uninstall` or `--stop` to manage background core and local server.
Agent Relay's Use Cases
- Automated Agent Handoffs
- Distributed Agent Workflows
- Real-time Agent Collaboration
- Conflict Resolution
- Event-Driven Agent Actions
- Secure Agent Communication
- Agent-to-Agent Querying
- Session History and Replay


