Kilntainers

Name: Kilntainers
Author: Kiln-AI

@Kiln-AI

リポジトリ

Kilntainers について

Kilntainers is an MCP server that gives LLM agents isolated Linux sandboxes for executing shell commands.

#containers #docker #podman #modal #e2b

基本情報

カテゴリ

その他

トランスポート

stdio

公開者

Kiln-AI

投稿者

Steve Cosman

設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "kilntainers": {
      "command": "kilntainers"
    }
  }
}

ツール

ツールは検出されませんでした

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is Kilntainers?

Kilntainers is an MCP server that gives LLM agents isolated Linux sandboxes for executing shell commands. It supports multiple backends including local containers (Docker, Podman), cloud-hosted micro-VMs (Modal, E2B), and WebAssembly sandboxes (WASM BusyBox or custom WASM modules).

How to use Kilntainers?

Install via uv tool install kilntainers (optionally with [wasm] for WASM backends). Run the server from CLI with kilntainers; by default it uses Docker and Debian slim. Add the server to an MCP client by specifying "command": "kilntainers" in the MCP configuration. The single MCP tool sandbox_exec lets the agent run any Linux command.

Key features of Kilntainers

Multiple backends: Docker, Podman, Modal, E2B, and WASM.
Isolated per agent – each MCP connection gets its own sandbox.
Ephemeral sandboxes – cleaned up automatically when the session ends.
Secure design – agent communicates with sandbox, not inside it.
Single MCP tool: sandbox_exec for running any command.
Scalable from laptop to thousands of parallel cloud sandboxes.

Use cases of Kilntainers

Letting agents use Linux utilities like grep, find, jq to save tokens.
Running thousands of parallel agents with isolated, ephemeral environments.
Safe command execution without exposing host OS or agent secrets.
GPU-accelerated sandboxes via Modal for compute-heavy agent tasks.

FAQ from Kilntainers

What backends are available?

Docker (or Podman), Modal, E2B, Go BusyBox WASM, and custom WASM modules. Each backend is selected via the --backend flag.

How does Kilntainers ensure security?

The agent communicates with the sandbox over MCP but does not run inside it. Agent secrets (API keys, prompts, code) are never exposed to the sandbox, preventing exfiltration via prompt injection.

What are the runtime requirements?

Python 3.13+ is required. The Docker backend needs Docker or Podman installed. Modal and E2B backends require accounts and authentication with those services. WASM backends need the optional [wasm]] dependency (+15 MB).

What transports does Kilntainers support?

Two MCP transports: stdio (default) and http (streaming HTTP mode). In HTTP mode, a single server can host many sandboxes in parallel.

How do I authenticate with cloud backends?

Modal: use modal setup CLI or provide --modal-token-id/--modal-token-secret flags. E2B: provide --e2b-api-key flag or the E2B_API_KEY environment variable.