Skills + Search2o

Search2o gives an Agent Skills client like Claude Code a controlled way to use your company's systems. This section calls these clients "the assistant". The assistant chooses which agent runs; it does not write what runs.

Why an agent, not code

In the assistant, the model can write code at run time and execute it with the person's access. A skill guides but does not define what will run. That works for resources the person chooses to expose to the assistant. It is a poor boundary for company systems.

Search2o gives the assistant access to published agents instead. An agent is a fixed definition, written by a company developer, validated by execution before publication, versioned, and unchanged at run time. The agent runs on a server inside your organization, within a runtime that limits what it can do. Code the assistant writes locally cannot access the systems behind the agents.

The assistant sees the request and the agent's output. Everything in between — API responses, database rows, credentials and agent variables — stays inside Search2o. Conversation state is encrypted and saved for follow-up requests. Every run is recorded by agent, version and person.

Together, Agent Skills provide the interface and Search2o provides the controlled execution behind it. People work with the assistant they already use, while the company defines what its agents can do and which systems they can access.