Content Router MCP

What it is

Same material, written once for one channel, then rewritten from scratch for three others. Yet the more expensive mistake is a different kind of fatigue: spending two days turning a long-form piece into a 60-second video that never had any narrative tension, while a throwaway case note that was born to be an infographic gets forced into an essay.

The problem is not execution — it is not knowing which format fits. Guess, and half your effort goes into the wrong shape.

Other servers: Content Router MCP · Memory Dream Engine MCP · Content Pattern Radar MCP · GEO MCP · China AI Compliance MCP · AI Customer Service MCP · All MCP servers

Deployment

Zero third-party dependencies · Pure Python standard library · Read-only

What it returns

• An 11-dimension profile — length, information density, structure type, visual potential, depth, emotional tone, narrative tension, visual-storytelling potential, infographic potential, actionability, timeliness
• Format recommendations with scores and reasons — each candidate carries a fit score = share of criteria met, comparable across formats rather than an arbitrary ranking
• A dispatch prompt — a ready-to-hand prompt for the format you pick, executable by any agent
• A batch decision table — point it at a whole material folder and get "do these first, in these formats"

When the top recommendation scores below 40%, it says outright: "this piece has no obvious format yet — do not start yet." It refuses to push, which is why it can be trusted.

Tools (5, all read-only)

Purely computational and read-only: all four annotation hints declared, no writes, no external requests, no API key.

Six cross-medium output formats

"Cross-medium" means the medium itself changes — text becomes a chart, a video, a comic, slides, or a reusable skill.

Platform rewriting (the same article for a different channel) is same-medium, different-platform and out of scope here — route first, rewrite second, so the two do not fight each other.

How to use it

This is an open-source, locally deployed MCP (pure Python standard library, no mandatory dependencies). Clone it and wire it into your client:

git clone https://github.com/savantcat/content-router.git
cd content-router
python server.py                  # stdio (Claude Desktop / Cursor / any MCP client)
python server.py --transport http --port 8769   # self-hosted streamable-http
python server.py --selftest       # bypass the protocol and hit every tool

Client configuration (stdio):

{
  "mcpServers": {
    "content-router": { "command": "python", "args": ["server.py"] }
  }
}

Published to the official MCP Registry as cn.savantcat/content-router. You can skip MCP entirely and use the CLI: python engine.py batch ./materials --top 20 --out plan.md.

Honest boundaries

• The profile can be wrong. You can correct dimensions by hand and re-run the recommendation.
• It judges format, not quality. "This suits an infographic" says nothing about whether the writing is any good.
• Not a substitute for your judgement. The fit score is a share of criteria met, not a content-quality score.
• Zero dependencies, pure computation. No network, no LLM calls, no API key.
• Method published. profile_schema returns every definition and range so you can audit the boundaries rather than trust a number.

Citation and licensing

This server is open source (MIT) and read-only; agents and developers are welcome to call it. When citing our measurements, credit 合尘猫 SavantCat (savantcat.cn).