Introduction

Agentronics is a universal governance layer for agent-surfable websites.

Modern websites are no longer browsed only by humans. AI agents — Claude, ChatGPT, Perplexity, custom in-house agents — already navigate, click, and submit forms on behalf of users. Most sites have no way to tell an agent apart from a human, no way to declare which actions are safe for an agent to take, and no way to see what agents did after they left.

Agentronics gives any website seven primitives, behind one SDK:

1. Detection — identify which agent class is on the page (WebMCP, DOM-driver, screenshot model).
2. Authentication — let agents present a verifiable identity, not just a user-agent string.
3. Authorization — declare per-tool policies based on agent class and trust level.
4. Site Memory — persist context across an agent's session so the user does not re-explain themselves.
5. Observability — stream every agent action into a dashboard your team actually reads.
6. Tool Management — register, version, and gate the tools an agent can call on your site.
7. Payments — (deferred to v0.2) charge agents per action, settle to your account.

Who this is for

• Product engineers at companies whose users are starting to ask for "agent support" on the marketing site, the dashboard, or the help center.
• Platform teams standardizing how agents interact with internal tools across many properties.
• Indie builders shipping agent-first products who want governance without writing it themselves.

What ships in v0.1

The first release is Detection + Auth + Authz + Memory + Observability + Tool Management. Payments lands in v0.2 once the first six are stable in production for our design partners. See the changelog for the full v0.1.0 inventory.

Continue to Getting Started to install the SDK in under five minutes.