AuroraScript.Mcp 3.0.0

AuroraScript MCP Server

This project provides a stdio MCP server for AuroraScript.

It intentionally keeps dependencies small and uses JSON-RPC over stdin/stdout.

Build

dotnet build language-tools/AuroraScript.Mcp/AuroraScript.Mcp.csproj

Pack dotnet tool

dotnet pack language-tools/AuroraScript.Mcp/AuroraScript.Mcp.csproj -c Release -o artifacts/mcp-tool

The package id is AuroraScript.Mcp and the tool command is aurora-mcp.

Local install smoke test:

dotnet tool install AuroraScript.Mcp --version 3.0.0 --add-source artifacts/mcp-tool --tool-path artifacts/mcp-tool-install

Publish to NuGet:

dotnet nuget push artifacts/mcp-tool/AuroraScript.Mcp.win-x64.3.0.0.nupkg --api-key <api-key> --source https://api.nuget.org/v3/index.json
dotnet nuget push artifacts/mcp-tool/AuroraScript.Mcp.linux-x64.3.0.0.nupkg --api-key <api-key> --source https://api.nuget.org/v3/index.json
dotnet nuget push artifacts/mcp-tool/AuroraScript.Mcp.osx-x64.3.0.0.nupkg --api-key <api-key> --source https://api.nuget.org/v3/index.json
dotnet nuget push artifacts/mcp-tool/AuroraScript.Mcp.osx-arm64.3.0.0.nupkg --api-key <api-key> --source https://api.nuget.org/v3/index.json
dotnet nuget push artifacts/mcp-tool/AuroraScript.Mcp.3.0.0.nupkg --api-key <api-key> --source https://api.nuget.org/v3/index.json

Users can install it with:

dotnet tool install --global AuroraScript.Mcp

Publish single-file executable

Publish with a runtime identifier to create a self-contained single-file MCP server that does not require a local .NET runtime:

dotnet publish language-tools/AuroraScript.Mcp/AuroraScript.Mcp.csproj -c Release -r win-x64 --self-contained true -o artifacts/mcp-win-x64-single

The documents language pack is embedded in the executable. The loose documents folder is only used as a development fallback.

Run

dotnet run --project language-tools/AuroraScript.Mcp/AuroraScript.Mcp.csproj

Run the published executable directly:

artifacts/mcp-win-x64-single/AuroraScript.Mcp.exe

PowerShell smoke test:

$exe = ".\artifacts\mcp-win-x64-single\AuroraScript.Mcp.exe"
'{"jsonrpc":"2.0","id":1,"method":"resources/read","params":{"uri":"aurora://schema/runtime-api"}}' | & $exe
'{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"aurora_get_document","arguments":{"id":"host-integration"}}}' | & $exe

Tools

• aurora_get_document: read a language-pack document by id.
• aurora_list_documents: list all embedded language-pack documents, schemas, and examples.
• aurora_list_features: return structured language feature metadata.
• aurora_check_script: compile-check a module or block and return diagnostics. Supports in-memory dependency sources.
• aurora_run_script: compile and run a module or block, returning result, stdout, stderr, diagnostics, and runtime errors.
• aurora_check_file: compile-check a file-system entry .as file and its import/include graph.
• aurora_run_file: compile and run a file-system entry .as file and its import/include graph.
• aurora_build_workspace: compile all .as files visible under a file-system root.
• aurora_search_runtime_api: search script-side runtime APIs such as String.trim, HashMap, or appendLine.
• aurora_get_runtime_api: read one runtime API entry by path such as Array.push.
• aurora_list_examples: list valid and invalid examples from the embedded manifest.
• aurora_get_example: read an embedded example source file.
• aurora_validate_best_practices: check generated script for AI authoring warnings.
• aurora_explain_diagnostic: explain common AuroraScript diagnostics.

Resources

• aurora://docs/ai
• aurora://docs/script-best-practices
• aurora://docs/language
• aurora://docs/performance
• aurora://docs/host-integration
• aurora://schema/ebnf
• aurora://schema/features
• aurora://schema/runtime-api
• aurora://schema/host-api
• aurora://examples/manifest

Tool Example

Read AI authoring guidance:

'{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"aurora_get_document","arguments":{"id":"script-best-practices"}}}' | & $exe

{
  "name": "aurora_check_script",
  "arguments": {
    "mode": "module",
    "sourceName": "main.as",
    "source": "@module(TEST); import lib from './lib'; export func run() { return lib.value(); }",
    "sources": {
      "lib.as": "@module(LIB); export func value() { return 42; }"
    }
  }
}

Run a script and capture output:

{
  "name": "aurora_run_script",
  "arguments": {
    "mode": "module",
    "source": "@module(TEST); export func run() { console.log('value', 42); return 42; }"
  }
}

Run a CompileBlock body with parameters:

{
  "name": "aurora_run_script",
  "arguments": {
    "mode": "block",
    "source": "return left + right;",
    "parameters": ["left", "right"],
    "arguments": [20, 22]
  }
}

Run a real .as file from disk and let the file-system resolver load imports/includes:

{
  "name": "aurora_run_file",
  "arguments": {
    "rootDirectory": "examples/tests",
    "entryPath": "main.as",
    "moduleName": "MAIN",
    "methodName": "run"
  }
}

Check a file with an in-memory overlay that overrides or adds sources before disk lookup:

{
  "name": "aurora_check_file",
  "arguments": {
    "rootDirectory": "scripts",
    "entryPath": "main.as",
    "sources": {
      "generated/config.as": "@module(CONFIG); export const value = 42;"
    }
  }
}

Search runtime APIs before generating code:

{
  "name": "aurora_search_runtime_api",
  "arguments": {
    "query": "StringBuffer",
    "limit": 10
  }
}

Validate AI authoring guidance:

{
  "name": "aurora_validate_best_practices",
  "arguments": {
    "mode": "block",
    "source": "for (var i = 0; i < items.length; i++) { total += items[i]; } return total;"
  }
}