Skip to main content
Glama
deenesjakoruzh
vigilancetrent

mcpforge

by vigilancetrent
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Hybrid

mcpforge

5行のPythonコードでMCPサーバーを構築。

Python License: MIT Status MCP

Before                              After
──────                              ─────
~200 lines of                       @serve
JSON-RPC stdio                      class MarketTools:
boilerplate, schema                     @tool
generation, error                       def latest_price(self, symbol: str) -> float: ...
handling, lifecycle                     @tool
management.                             def search(self, query: str) -> list[dict]: ...

                                    $ python -m mcpforge run market:MarketTools
                                    ✓ MCP server running on stdio

mcpforgeを選ぶ理由

Model Context Protocol (MCP) は、ClaudeやCursor、そして急速に拡大するAIツールエコシステムが外部関数を呼び出すために使用するオープン標準です。強力ですが、サーバーを書くのは現在非常に面倒です。

プロジェクトごとに同じJSON-RPCのフレームワーク、スキーマ生成、ディスパッチループ、エラーハンドリングを書く必要があります。mcpforgeはこれらすべてを不要にします。 クラスにアノテーションを付け、メソッドに型ヒントを付けるだけです。これで完了です。

from mcpforge import serve, tool

@serve(name="market_tools", version="0.1.0")
class MarketTools:
    """Market data tools for AI agents."""

    @tool(description="Get the latest price for a symbol")
    def latest_price(self, symbol: str) -> float:
        return fetch_price(symbol)

    @tool
    def search(self, query: str, limit: int = 10) -> list[dict]:
        """Search ticker symbols matching `query`."""
        return run_search(query, limit)

これで、仕様に完全に準拠したMCPサーバーが完成します。型ヒントはJSON Schemaになり、ドキュメント文字列はツール説明になり、戻り値は自動的にシリアライズされます。

Related MCP server: nbmcp

60秒クイックスタート

pip install mcpforge
# hello.py
from mcpforge import serve, tool

@serve(name="hello", version="0.1.0")
class HelloTools:
    @tool
    def greet(self, name: str = "world") -> str:
        """Say hello to someone."""
        return f"Hello, {name}!"
python -m mcpforge run hello:HelloTools

これで、stdio経由でJSON-RPC 2.0を話す動作可能なMCPサーバーが完成しました。

仕組み

  1. @serve はクラスをMCPサーバー（名前、バージョン、機能）としてタグ付けします。

  2. @tool はメソッドをMCPツールとして登録します。

  3. @resource(uri=...) はメソッドをMCPリソースとして登録します。

  4. mcpforgeは各メソッドのシグネチャをイントロスペクトし、JSON Schemaを生成します。プリミティブ、list[T]、dict[str, T]、Literal[...]、Optional[T]、データクラス、Pydanticモデルがすべてそのまま動作します。

  5. python -m mcpforge run mod:Class はstdioループを起動し、initialize / tools/list / tools/call / resources/list / resources/read を処理し、適切なコードでJSON-RPCエラーを報告します。

外部のMCP SDKは不要です。標準ライブラリの json と、スキーマを便利にするための pydantic だけを使用します。

組み込みサーバー

今日からすぐに使える2つのサーバーが同梱されています：

# Filesystem tools (sandboxed to a root directory)
python -m mcpforge run mcpforge.builtin.filesystem:FilesystemTools

# HTTP fetch tools
python -m mcpforge run mcpforge.builtin.http:HttpTools

FilesystemTools は list_dir、read_file、search を公開し、パストラバーサルチェックでサンドボックス化されています。HttpTools はサイズ制限付きの fetch_url を公開します。

Claude Desktopへの組み込み

~/Library/Application Support/Claude/claude_desktop_config.json に追加します：

{
  "mcpServers": {
    "market": {
      "command": "python",
      "args": ["-m", "mcpforge", "run", "market:MarketTools"],
      "cwd": "/path/to/your/project"
    }
  }
}

Claude Desktopを再起動してください。会話の中にツールが表示されます。

Cursorへの組み込み

~/.cursor/mcp.json：

{
  "mcpServers": {
    "market": {
      "command": "python",
      "args": ["-m", "mcpforge", "run", "market:MarketTools"]
    }
  }
}

比較

機能

mcpforge

mcp SDK

FastMCP

手書き

単一デコレータ

Yes

No

Yes

No

型からの自動JSON Schema生成

Yes

No

Yes

No

Pydantic v2サポート

Yes

Yes

Yes

No

pyd以外の必須依存関係ゼロ

Yes

No

No

N/A

組み込みfs / httpサーバー

Yes

No

No

No

Hello Worldの行数

~5

~40

~10

~200

ワイヤーフォーマットの確認

python -m mcpforge inspect market:MarketTools

クライアントが確認する正確な tools/list ペイロードを出力します。

ロードマップ

  • [x] ツール (call, list)

  • [x] リソース (read, list)

  • [x] 型ヒント -> JSON Schema (Pydantic v2, dataclasses, Literal, Optional)

  • [x] 組み込みファイルシステムおよびHTTPサーバー

  • [ ] リソースサブスクリプション

  • [ ] プロンプト機能

  • [ ] サンプリング機能

  • [ ] WebSocket / HTTP-SSEトランスポート (オプションの [http] エクストラ)

  • [ ] 非同期ツール (async def)

  • [ ] OTelトレーシングフック

ライセンス

MIT — LICENSE を参照してください。

作者: thechifura。quantflow および strategos の兄弟プロジェクト。

Available Tools

3 tools
list_dirA

List entries in path (relative to the sandbox root).

ParametersJSON Schema
NameRequiredDescriptionDefault
pathNo.

TDQS

A3.5/5.0
Behavior2/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

No annotations provided, so description bears full burden. It does not disclose what types of entries are listed (files, directories), sorting order, or any other behavioral traits beyond listing.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Single sentence with no wasted words. Verb is front-loaded, and the essential context is included efficiently.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness3/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a simple listing tool with one parameter and no output schema, the description is adequate but could be more complete by noting whether entries include both files and directories.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Input schema has 0% description coverage, but description adds meaning by specifying that path is relative to the sandbox root. This clarifies a key detail not in the schema.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Description clearly states the verb 'list', the resource 'entries in path', and adds context 'relative to the sandbox root'. It distinguishes from sibling tools read_file and search.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines2/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

No guidance on when to use this tool versus alternatives or exclusions. The description does not mention any context or conditions for usage.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

read_fileA

Read up to max_bytes bytes of the file at path as UTF-8 text.

ParametersJSON Schema
NameRequiredDescriptionDefault
pathYes
max_bytesNo

TDQS

A3.7/5.0
Behavior3/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

The description discloses that it reads up to max_bytes as UTF-8 text, but does not mention error behavior (e.g., file not found, binary files) or performance implications. With no annotations, the description provides moderate transparency.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single, clear sentence that efficiently covers the tool's core functionality without extraneous words.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness3/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given no output schema and sibling tools, the description could mention return value format or error handling. It adequately covers the main behavior but lacks completeness for a file reading tool.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 0%, so the description must compensate. It explains both parameters: path is the file location, max_bytes limits the bytes read, and specifies UTF-8 encoding. However, it does not clarify path format (absolute/relative).

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states it reads a file at a given path, up to a specified number of bytes, as UTF-8 text. It distinguishes from sibling tools 'list_dir' (listing directory contents) and 'search' (searching for files).

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines2/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

No guidance is provided on when to use this tool versus alternatives. There is no mention of when not to use it or any prerequisites.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

  1. 3 tool updatesv0.1.0
    • First observedlist_dir
    • First observedread_file
    • First observedsearch
Deploy Server
Shareable, User-Owned Context BackendPeople, applications, and agents read, write, and coordinate together in one context lake. Reach it over MCP, a CLI, a Python SDK, or REST.fulcradynamics.comAd

TDQS

B3.3/5.0

Scored across 3 tools

Disambiguation5/5

Each tool has a clearly distinct purpose: listing directory entries, reading file contents, and searching for patterns. No overlap in functionality.

Naming Consistency4/5

Naming is mainly verb_noun (list_dir, read_file) with one tool using a single verb (search). All use consistent snake_case, but slight pattern deviation prevents a perfect score.

Tool Count5/5

Three tools is a reasonable and focused set for basic file operations within a sandbox. Each tool serves a distinct and necessary function.

Completeness2/5

Only read-only operations are provided. Missing write, update, or delete tools, which are essential for a file manipulation server, leaving significant gaps.

Maintenance

ActivityInactive
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers

  • cleanmcp

    Lishang25
    F
    license
    Not graded
    quality
    D
    maintenance
    A MCP server framework with zero-config auto-discovery, type-safe decorators, and HTTP transport, enabling easy creation and deployment of tools, prompts, and resources to LeanMCP Cloud.
    -

  • nbmcp

    rum1887
    A
    license
    Not graded
    quality
    D
    maintenance
    Enables building lightweight MCP servers where Rust handles transport, routing, and schema validation while Python defines tool behavior using type hints, with support for stdio and HTTP transports.
    MIT

  • mcp-forge

    benni-os
    A
    license
    Not graded
    quality
    B
    maintenance
    A FastAPI-style framework for building MCP servers in Python, enabling declarative tool definitions with automatic schema validation and multiple transport options.
    MIT

  • nn-mcp-stdio

    fre-sch
    A
    license
    Not graded
    quality
    B
    maintenance
    A no-nonsense framework for building stdio MCP servers using plain async handlers and type annotations, with strict schema validation and no configuration overhead.
    MIT