Interaction layer between humans and agents

Software is becoming dynamic.

Software as Content (SaC) shapes AI responses into dynamic apps that are intuitively readable and interactive. It structuralizes prompts to guide AI, and keeps evolving across every interaction and conversation.

The paradigm

GUIs made software manipulable.
SaC does the same for AI.

GUIs turned command lines into something anyone could use. SaC does the same for AI — turning chat replies into software you can see, click, and evolve.

Past
CLI
type commands, read output
GUI
see & manipulate directly
Now
Chat
ask in text, read replies
SaC
interact with living software

Live demos

One prompt. A living app.

Type what you need, get an interactive app you can click through and keep evolving. Try it live below.

Loading live app…

Open source

SaC open-sources its framework.
Connect to your agent in minutes.

Three integration modes — MCP, Skill, or Python. Same loop: generate an app, receive user actions, evolve or reply.

Check Open-source SDK
Loading video…

In the wild

Works with mainstream agents
out of the box.

Claude Code, Codex, OpenClaw — plug SaC in and your agent starts generating interactive apps users can click through.

Architecture visualization

Architecture visualization

Claude Code reads a codebase and generates an interactive architecture diagram — clickable nodes, animated data flows, dependency risk analysis.

Check Open-source SDK

The SDK

Connect your agent in one line.

Three integration modes. Same interaction loop: generate an app, receive user actions, evolve or reply.

$ pip install sac-sdk
$ sac setup claude-code

# MCP tools available to the agent
generate_app(intent)
evolve_app(conversation_id, intent)
send_chat(conversation_id, message)
wait_for_action(conversation_id)

The next form
of software.

Try SaC live, or bring the open-source SDK to your own agent.

View on GitHub
Or join the waitlist for updates