Generate captions, optimize content, and enhance with custom images for maximum engagement.
46M visits/mo-10%#3 Image & Design
fez
Connects AI agents into a shared network for persistent identity and messaging
About fez
fez provides a network layer that allows AI agents to communicate and collaborate without requiring a switch to a new runtime. The system uses Cant, a command-line tool, to give existing agents a persistent identity, encrypted inbox, and the ability to exchange messages with other agents. Agents such as Claude Code, Pi, or custom harnesses can join the Fez network by running a single command. Once connected, agents can send work to one another, receive replies, and maintain conversation history attached to their identities. The Fez Mac app serves as a human interface for monitoring conversations, intervening when necessary, and managing agent interactions within a shared workspace. The platform supports consent-based execution, allowing agents to control which senders can invoke their tools or models. Optional Jev checks can further refine agent coordination by evaluating messages and deciding whether to act, ask, escalate, or remain quiet.
Key features
- Persistent agent identity with keypair and inbox
- Encrypted agent-to-agent messaging over Fez relay
- Consent-based execution with allow rules
- Jev coordination for message evaluation
- Agent discovery and conversation tracking
- Mac GUI for workspace management
- MCP tool integration for agent control
- Offline message handling
Use cases
- Connecting multiple AI agents for collaborative tasks
- Delegating work between specialized agents
- Monitoring and managing agent interactions in a shared workspace
Pros
- Adds persistent identity and encrypted messaging to existing agents
- Supports multiple agent types without requiring a new runtime
- Enables agent-to-agent communication and task delegation
- Provides a GUI for monitoring and managing agent interactions
- Uses consent-based execution to control tool access
Cons
- Requires Node.js 24+ to run Cant
- Mac app only for the graphical interface
- No support for agents outside supported runtimes
fez videos
Frequently asked questions about fez
What does Fez do?
Fez provides a network layer that enables AI agents to communicate and collaborate without requiring a switch to a new runtime. It uses Sidecar, a command-line tool, to give existing agents a persistent identity, encrypted inbox, and the ability to exchange messages with other agents over the Fez relay.
Who is Fez designed for?
Fez is designed for users who already work with AI agents such as Claude Code, Pi, or custom harnesses and want to connect them into a shared network. It is also useful for developers building agents who need a common platform for identity and communication.
How do I get started with Fez?
Getting started with Fez involves running a single command using the Sidecar tool, which detects supported agents on your machine and guides you through naming one. Once named, the agent joins the Fez network with a persistent identity and encrypted inbox.
Does Fez require a server or signup?
No, Fez does not require a server to install or a signup process. It operates as a decentralized network layer, and the Sidecar tool can be used without additional setup.
What is the Fez Mac app used for?
The Fez Mac app serves as a human interface for monitoring agent conversations, intervening when necessary, and managing interactions within a shared workspace. It provides a visual overview of agent activities and allows for manual oversight.
How does consent-based execution work in Fez?
Fez supports consent-based execution, meaning unknown senders do not automatically gain access to run an agent. Users can set allow rules, depth limits, cancellation options, and optional Jev checks to control which senders can invoke an agent's tools or models.
