Website: search2o.com · Source: GitHub · GUI source: GitHub · Docs: search2o.com/docs . Skill Search2o-skill

This repository contains the Search2o Agent Server: a stateless, asynchronous Python server that runs agents inside your organization, with a REST API and a bundled GUI. Search2o is in open beta.

Getting started

1. Create your account

Create your account at Getting started. Your license key appears on that page when the account is created. No email is sent at this step.

2. Install the agent server

Install the agent server with the bundled GUI:

pip install search2o

To install from the source instead, clone the agent server and install it from the clone. The clone does not include the GUI:

git clone https://github.com/Search2o/agent-server.git
pip install -e agent-server

To install the GUI from the source, follow the instructions in the README of the GUI repository.

3. Run the server

macOS / Linux:

SEARCH2O_LICENSE_KEY=your-license-key search2o

Windows PowerShell:

$env:SEARCH2O_LICENSE_KEY="your-license-key"
search2o

Self-hosted models using Ollama, vLLM or SGLang need no vendor API key. For OpenAI, Anthropic or Gemini, before starting the server, also set one or more of OPENAI_API_KEY, ANTHROPIC_API_KEY or GEMINI_API_KEY. For other LLMs, see Connecting to other LLMs.

4. Set up your account

Open http://127.0.0.1:9020/ui and click New user / Forgot password on the sign-in dialog. Enter the same email you used to create your account in step 1. A code is emailed to that address and no other. Enter the code, choose a password, and sign in.

The server also serves its OpenAPI schema at /openapi.json, Swagger UI at /docs and ReDoc at /redoc.

5. Build your first agents

LLMs are called from an agent using an LLM profile. You can use an existing profile or create a new one.

1. Open Agents → Drafts and create a draft. Select your LLM profile, then tell Draft with AI what you want.
2. Build two or three task-specific agents. Each agent would typically use one or two of your enterprise systems and handle a specific use case.
3. Validate and publish them. Then add plain-English descriptions so search can find them.
4. Type a question in the search box: the agent matches, runs and answers.

The same agents work in Claude Code through the search2o-skill.

If you have questions, click the docs icon at the top of the UI, or see the full documentation at search2o.com/docs. We would love to hear how setting up and running your first few agents went: email info@search2o.com, or send a message from the message icon in the GUI.

License

The Search2o Agent Server is source available and proprietary. Use is subject to the Search2o Software License Agreement.

Contributions

External code contributions and pull requests are not currently accepted. Bug reports, feature requests, documentation feedback, and security reports are welcome. See CONTRIBUTING.md.