# An agent server with a framework, development environment, runtime, semantic router, and state management — retained record

> Retained reddit record from reddit:AI_Agents.

Canonical: https://highsignal.app/data/records/0cf0f99fa33f05db

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:AI_Agents)  published 2026-11-03 15:40 ingested 2026-11-04 02:36

[original source ↗](https://www.reddit.com/r/AI_Agents/comments/1xmwa4j/an_agent_server_with_a_framework_development/)

## retained text



```
I built Search2o and would appreciate your feedback before the official launch.

Search2o is an agent server that includes an agent framework, a development environment, a runtime, a semantic router, and cloud-based state management.

**Agent framework:** Agents are defined in JSON with Python expressions. Twenty commands support agentic workflows, including tool calling, long-term memory, and human-in-the-loop interactions. AI can draft agents from a description. The development environment includes a schema-aware editor, tracing, validation, and publishing.

**Agent runtime:** Runs agents in a controlled environment that restricts what Python expressions can do and enforces limits on execution time and LLM cost.

**Search (semantic router):** Developers describe agents in plain English. Search matches a user’s request against those descriptions and returns the best agent in about half a second. The client runs the matched agent on the server with the same request. Users don’t need to know which agents exist, so the system can grow to hundreds of agents without changing how they use it.

**Conversation and state management:** State, including short-term memory and prompts, is encrypted and saved in the cloud. A follow-up request in the same conversation can go to another agent, which uses the same context and builds on previous results.

**Model neutrality:** The agent server talks to LLMs directly, using your own API keys where needed. Adapters are included for OpenAI and compatible APIs, Claude, Gemini, Ollama, vLLM, and SGLang. You can also create your own adapters for other LLMs.

You can try several live demos without signing up. I will post the link in the comments. 

If you’re involved in enterprise AI, I’d appreciate your thoughts: what looks useful, and what is missing?
```



document fetched 2026-11-04 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1xmwa4j/an_agent_server_with_a_framework_development
