What is Search2o

Search2o is a platform for building AI agents and running them from a search interface. A user types what they need; Search2o finds the agent built for that need, runs the agent, and returns the answer in a conversation the user can continue.

Developers write agents in a declarative JSON format with Python expressions, and describe each agent in plain English. Search2o matches a user's query to the agent whose description fits, runs that agent in a controlled runtime, and returns the response. A follow-up may go to the same agent or to a different agent; the conversation carries its context along.

Search2o is self-contained. Create an account and install the agent server. The search box, the code editor, user management and reports are all there; nothing has to be built around Search2o.

What makes it different

• Search is the orchestration layer. An organization accumulates many focused agents, written by different teams, and search makes the whole collection behave like one application.
• Agents are structured and declarative. An agent is a set of functions, and a function is an ordered list of commands. The 23 commands cover what agents are built to do: call APIs and databases, prompt LLMs with tools, ask the user for input, and remember facts across conversations.
• Execution is controlled. The Python expressions in an agent run in a controlled runtime, governed by the allowlist, the compile rules and the runtime limits that developers set.
• State is managed. The agent server holds no state - everything is stored in Search2o cloud, so servers can be started, stopped and scaled without concerning about data loss. Sensitive information like conversation, queries, and memories are encrypted.