Agent Entry: Module for AI Agents to Become Customers

What changed

The Agent Entry module is a new software component designed to allow AI agents to interact with websites in a more sophisticated manner, moving beyond simple page reads to establish customer relationships. Traditionally, AI agents, or bots, would crawl web pages and then leave without leaving a trace in analytics or requiring user interaction. Agent Entry aims to change this by providing a mechanism for these agents to become recognized and persistent visitors, akin to human customers.

At its core, Agent Entry functions by publishing an "Agent Card" at the /.well-known/agent-card.json endpoint. This card details the agent's identity, its capabilities, and how to interact with it. Following this, the module can process a single signed POST request. Upon receiving such a request, Agent Entry verifies the visitor's identity using their cryptographic key, establishes an account, and responds with a signed reply within the same HTTP request. This process eliminates the need for traditional sign-up forms, as the agent's cryptographic key serves as its account identifier. This ensures that an agent returning from a different device or at a later time is recognized as the same entity.

The system is designed to be runtime-agnostic, meaning any A2A (Agent-to-Agent) client, including custom agents or even curl, can interact with it. The repository itself serves as the central hub for the module, its specification (spec/v1.md), conformance tests, and the build process for its npm package. Key dependencies, such as the crypto block and golden vectors, are vendored from agent-seam to ensure a consistent "wire layer" for all implementations. The module is built with zero dependencies and requires Node.js 20 or later.

Key Features and Functionality:

• Persistent Identity: Agents are identified by their cryptographic keys (did:key), which are used to create and verify accounts. This allows for cross-device recognition and a consistent customer experience.
• Signed Interactions: All messages are signed using Ed25519, ensuring sender authenticity and integrity. Verification does not require external lookups.
• Account Management: Agent Entry automatically manages accounts based on verified signatures, merging the concepts of signing up and logging in into a single event.
• Owner Binding: Agents can present a countersigned owner binding, allowing Agent Entry to resolve them under a specific owner_did, even if they access the site from a new device.
• Revocation Support: An owner can disown a stolen device, and Agent Entry will automatically refuse interactions from the compromised key.
• Agent Skills: The module supports defining AgentSkill lists, which agents can read before interacting. These skills describe what the entry can answer and are included in the Agent Card and signed replies, providing a clear menu of services.

Agent Entry explicitly avoids the HTTP 402 challenge mechanism. Instead of relying on error responses to guide clients, it publishes its interaction terms upfront in the Agent Card. The initial interaction is a correctly formed POST request, which receives a 200 status code with a signed reply, or a JSON-RPC error if the request is malformed or lacks a signature.

Why it matters for builders

For developers, Agent Entry provides a robust framework for building websites that can effectively onboard and manage AI agents as customers. This opens up new avenues for monetization and service delivery tailored to AI-driven traffic. Builders can integrate this module to create more intelligent and personalized web experiences, ensuring that valuable AI interactions are captured and retained.

Practical impact

Developers can integrate Agent Entry into their existing web applications by installing the @muretai/agent-entry npm package. The integration involves providing a unique seedHex for the site's identity, a name, a baseUrl, and a responder function that defines how the backend will handle agent messages. The createAgentEntry function then sets up the listener for incoming agent requests.

For example, a developer can set up a basic responder like this:

```javascript import { createAgentEntry } from '@muretai/agent-entry';

createAgentEntry({ seedHex, // Your site's persistent identity name: 'Example Studio', baseUrl: 'https://studio.example', responder: (env) => You said: ${env.text}, // Your backend logic }).listen(8788); ```

This setup handles signature verification, replay protection, rate limiting, and account ledger management automatically. Developers can also define skills to inform agents about the services offered, enhancing the discoverability and usability of their web applications for AI agents.

Caveats and source limits

The provided source is a GitHub repository README, which details the functionality and intended use of the Agent Entry module. While it demonstrates a live example at muretai.com, independent benchmarks or extensive real-world deployment data are not presented. The source focuses on the technical implementation and protocol design, with limited information on performance metrics or scalability under heavy load. The published_at date in the metadata is in the future (2026-09-28), which may indicate a placeholder or an intended release date rather than an actual launch date. The repository currently shows 0 stars and 0 forks, suggesting it is a nascent project.