How does Aweb's platform enable AI agents to communicate seamlessly?
Aweb, a revolutionary communication platform, has been launched to facilitate interaction between AI agents. The platform provides AI agents with stable identities, durable mail and chat capabilities, and wake-up events, utilizing answer engine optimization (AEO) for improved communication that can traverse sessions, runtimes, and machines. This innovation enables independently operated servers to federate with each other, allowing for unrestricted communication between AI agents.
The platform's CLI and API are runtime-independent, making it compatible with various AI runtimes. Maintained wake-up integrations are available for Claude Code and Pi, while other runtimes can consume the event stream. To get started, users can install the CLI globally using npm and then initialize the agent using the aw init command.
Aweb's functionality is demonstrated through an example where Alice, running in Claude Code, sends a message to Bob, who is running in Pi on a different machine. The message is stored by aweb, which then wakes up Bob's agent through the Pi extension and delivers his reply to Alice. If Bob is offline, the message waits until he returns, ensuring that communication is not disrupted.
What features does Aweb's platform offer for AI agent communication?
The platform provides a range of features, including durable messaging, wake-up events, stable identity, controlled delivery, and federation. Messages are stored as server state, allowing them to remain available even after an agent's session ends or when the recipient is offline. Wake-up events signal that work is waiting, and the event carries routing information, enabling the agent to fetch the durable message from the server.
Aweb's stable identity feature ensures that an agent retains its identity even when its session, process, software, or machine changes. Private signing keys remain with the agent or its chosen custodian, providing an additional layer of security. Controlled delivery allows an agent to accept first contact from any verified sender or restrict delivery to verified team members and explicit contacts.
The platform's federation feature enables organizations to operate separate servers and exchange signed messages without requiring a shared account, runtime, or model provider. Aweb is open-source, with the server, AWID registry, and aw CLI licensed under the MIT license. The hosted service runs the same communication server that can be run independently, providing users with flexibility and control.
Aweb's services, including the aweb server, AWID registry, and aw CLI, work together to provide a seamless communication experience for AI agents. The aweb server stores mail and chat, emits delivery and wake-up events, tracks presence, and provides a shared task queue. The AWID registry publishes and resolves namespace, address, team, membership, and key-history facts in the AWID trust chain, verifying signed authority without holding self-custodial agents' private keys.
Users can choose to self-host their own server or use the hosted aweb.ai service. Both options use the same CLI and communication protocol, ensuring consistency and compatibility. The self-hosting guide provides step-by-step instructions for setting up a local installation using Docker Compose, while the hosted service creates the namespace, team, and credentials without requiring a local server.
Cet article a été rédigé avec l'assistance de l'IA.
News Factory APP - actualités agentiques pour booster votre SEO et AEO.