SaC — Software as Content
VerifiedLet agents reply with live, evolving interactive apps instead of text.
What is the SaC — Software as Content MCP server?
SaC plugs into existing agents through the MCP protocol, letting Claude Code or similar clients call it to produce dynamic interfaces for tasks like trip planning or data dashboards. Users interact by clicking or chatting; both inputs loop back to the agent so the UI updates in place.
The package also ships a standalone HTTP server and Python SDK, but the MCP integration specifically registers SaC as a tool-capable server inside Claude Code after running sac setup claude-code.
Install & connect
Add this to your MCP client config. Pick your client below and copy.
{
"mcpServers": {
"sac": {
"command": "uvx",
"args": [
"sac-sdk"
],
"env": {
"SAC_API_KEY": "<YOUR_SAC_API_KEY>",
"SAC_API_BASE": "<SAC_API_BASE>",
"SAC_MODEL": "<SAC_MODEL>",
"SAC_SEARCH_API_KEY": "<YOUR_SAC_SEARCH_API_KEY>"
}
}
}
}Package: sac-sdk (pypi)
Example prompts
Once connected, try asking your AI client:
Configuration
Environment variables this server needs.
| Variable | Required | Description |
|---|---|---|
SAC_API_KEYsecret | Yes | API key for the LLM provider (OpenRouter, Anthropic, OpenAI, etc.) |
SAC_API_BASE | No | Custom API endpoint for OpenAI-compatible providers |
SAC_MODEL | No | Override the default generation model |
SAC_SEARCH_API_KEYsecret | No | Tavily API key for web search in generated apps |
Security & permissions
Runs locally over stdio and needs the SAC_API_KEY, SAC_API_BASE, SAC_MODEL and SAC_SEARCH_API_KEY environment variables to call external LLM and search services.
What you can do with SaC — Software as Content
Trip planning
Generate an interactive itinerary that users can modify by clicking dates or budget sliders while the agent refines the plan live.
Data dashboards
Create explorable visualizations of analysis results that evolve when the user asks follow-up questions or filters data.
Decision aids
Build comparison tools or financial models that update instantly as the conversation uncovers new constraints or preferences.
How to use SaC — Software as Content
- 1pip install sac-sdk
- 2sac setup claude-code to register the MCP server
- 3Restart Claude Code
- 4Ask the agent to use SaC for interactive tasks
- 5Keep sac serve running if using the HTTP viewer
SaC — Software as Content: pros & cons
Pros
- +Single evolving URL preserves context across turns
- +Works with existing agents via standard MCP
- +Supports both UI clicks and chat in one loop
- +Open prompts and design system for customization
Cons
- –Alpha release (v0.1.2) with limited documentation
- –Requires separate API keys for LLM and search
- –Best suited for exploratory tasks, not simple Q&A
Frequently asked questions
No tools are documented in the current README; the MCP server mainly enables the generate-and-evolve loop.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…