When building an Agentic System, you’ll need an API to serve your Agents, a database to store session and vector data and an admin interface for testing and evaluation. You’ll also need cron jobs, alerting and data pipelines for ingestion and cleaning. This system would generally take a few months to build, we’re open-sourcing it for the community for free.

​

What are Workspaces?

Workspaces are standardized codebases for production Agentic Systems. They contain:

• A RestAPI (FastAPI) for serving Agents, Teams and Workflows.
• A streamlit application for testing — think of this as an admin interface.
• A postgres database for session and vector storage.

Workspaces are setup to run locally using docker and be easily deployed to AWS. They’re a fantastic starting point and exactly what we use for our customers. You’ll definitely need to customize them to fit your specific needs, but they’ll get you started much faster.

They contain years of learnings, available for free for the open-source community.

​

Here’s how they work

• Create your codebase using: ag ws create
• Run locally using docker: ag ws up
• Run on AWS: ag ws up prd:aws

We recommend starting with the agent-app template and taking it from there.

​

How we build Agentic Systems

When building Agents, we experiment locally till we achieve 6/10 quality. This helps us see quick results and get a rough idea of how our solution should look like in production.

Then, we start moving to a production environment and iterate from there. Here’s how we build production systems:

• Serve Agents, Teams and Workflows via a REST API (FastAPI).
• Use a streamlit application for debugging and testing. This streamlit app is generally used as an admin interface for the agentic system and shows all sorts of data.
• Monitor, evaluate and improve the implementation until we reach 9/10 quality.
• In parallel, we start integrating our front-end with the REST API above.

Having built 100s of such systems, we have a standard set of codebases we use and we call them Workspaces. They help us manage our Agentic System as code.