SaC — Software As Content
Let your agent reply with evolving live React mini-apps instead of markdown walls when you prototype dashboards, planners, or interactive specs.
Overview
ai.dynsoft/sac is a MCP server for the Build phase that lets agents respond with live, evolving interactive apps (Software as Content) instead of static text.
What is this MCP server?
- SaC (Software as Content): agents return live, persistent, interactive apps that evolve with the conversation
- pip install sac-sdk; sac serve defaults to http://localhost:18420 with a built-in demo agent loop
- Claude Code integration documented via MCP; Codex via Skill; Python for custom agent loops
- Apache 2.0 SDK—not screenshots or static HTML exports
- Dynsoft-hosted MCP name ai.dynsoft/sac for catalog registration alongside local serve workflow
- Default local server URL http://localhost:18420 documented in Quickstart
- PyPI package sac-sdk with Apache 2.0 license
- Three integration paths called out: Claude Code MCP, Codex Skill, Python custom agent
Community signal: 3 GitHub stars.
What problem does it solve?
Agent replies are walls of markdown and one-off code blocks when you need a UI you can click, explore, and steer in real time.
Who is it for?
Solo builders prototyping interactive flows, internal tools, or visual specs inside Claude Code with a local sac serve loop.
Skip if: Teams that only want API backends, cannot run Python 3 locally, or need production hosting without the SaC interaction model.
What do I get? / Deliverables
After SaC is installed and MCP is connected, conversations can produce persistent React apps that update as you and the agent iterate.
- Live evolving React mini-apps inside agent conversations
- MCP or Skill bridge from Claude Code/Codex into SaC
- Local demo loop without an external agent for first-time evaluation
Recommended MCP Servers
Journey fit
SaC is an interaction layer for agents while you are building and iterating on product UX, so the canonical shelf is Build agent-tooling rather than distribution or ops. MCP and Skill hooks exist specifically to extend Claude Code and other agents with a persistent UI channel—core agent-tooling for the solo builder stack.
How it compares
Agent UI interaction MCP/SDK, not a component library skill or a static site generator.
Common Questions / FAQ
Who is ai.dynsoft/sac for?
Developers using Claude Code, Codex, or custom Python agents who want live React UIs as first-class replies during product exploration.
When should I use ai.dynsoft/sac?
Use it in the Build phase when wireframes, planners, dashboards, or multi-step tools are faster to shape as interactive apps than as markdown specs.
How do I add ai.dynsoft/sac to my agent?
Install sac-sdk from PyPI, run sac serve (configure API key on first launch), then attach the documented Claude Code MCP configuration from the SaC README.