The AI ecosystem is fragmented not just by company, but by file format.
- JSON: Used by Claude Desktop. Strict, no comments allowed (usually), brackets everywhere.
- TOML: Used by Python tools, Cargo, and some CLI agents. Clean, but whitespace-sensitive in its own way.
- YAML: Used by Kubernetes-style tools and some Gemini CLIs. readable, but indentation hell.
The Conversion Headache
You found a great MCP server on GitHub. The README gives you the configuration snippet in JSON. But you are using a tool that requires TOML.
You have to manually translate:
JSON Source:
{
"server": "git",
"args": ["--verbose"]
}TOML Target:
[server]
name = "git"
args = ["--verbose"]It looks easy until you have nested objects, environment variables with special characters, or arrays of objects. One missing quote or bracket, and the tool fails to start.
Automatic Transpilation
This is one of the core features of Vibe Manager.
We maintain an internal object model of an MCP Configuration.
- When we read from Claude, we parse JSON → Internal Model.
- When we write to Codex CLI, we render Internal Model → TOML.
This means you never have to think about trailing commas or indentation levels again. You define the intent (“I want the Git server”), and we handle the syntax for every specific tool.