MCP Server logo

MCP Server

by declaw-ai

Runs agent commands and file operations inside Declaw sandboxes with configurable network, PII, and prompt-injection controls.

Install MCP Server

Pick your client, then add it in one click or paste the config.

STDIONode.js
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "mcp-server-112": {
      "command": "npx",
      "args": [
        "-y",
        "@declaw/mcp-server"
      ],
      "env": {
        "DECLAW_API_KEY": "YOUR_VALUE_HERE",
        "DECLAW_DOMAIN": "YOUR_VALUE_HERE"
      }
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

Values are written into the snippet above so you can paste it as is. They never leave your browser.

Automated check passed— started and listed 7 tools correctly (8d ago).

Overview

The declaw-ai/mcp-server MCP server connects MCP-compatible AI tools to Declaw secure sandboxes for isolated command execution and file management. It creates and destroys sandboxes, runs shell commands, reads and writes files, and lists files or active sandboxes. Sandbox creation supports security presets, outbound domain allowlists, and audit logging, with stricter modes adding PII scanning and prompt-injection defense. Reach for it when an agent needs to execute untrusted or generated code with configurable network and data-safety policies.

Use cases

  • Run generated code in an isolated sandbox
  • Restrict sandbox traffic to approved domains
  • Inspect and modify files during agent workflows
  • Apply prompt-injection and PII safeguards to execution
  • Clean up temporary agent sandboxes

Key features

  • Firecracker microVM sandbox execution
  • Configurable security presets
  • Shell command execution
  • Sandbox file operations
  • Network domain allowlists
  • PII scanning and prompt-injection defense

Tools7

What MCP Server lets your agent do. Expand a tool for its parameters.

Verified live~779 tokens of context

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

Not checked yetLast checked Today

We have not completed a health check for this listing yet.

8/16 checks healthy over the last 47d

What it can do

From the 7 tools this server returned when we ran it.

Runs commands · 1Can delete or modify data · 1Writes files · 1Handles credentials · 1Read-only tools · 4
Network at startup3 connections
Files written at startupNone
Install scriptsprepare
npm provenanceSigned build
Package age138 days
Maintainers1

Observed 2026-10-21 in an isolated sandbox. A description of reach, not a security audit.

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 6d ago via OSV.dev · @declaw/mcp-server (npm)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars1Stargazers on the source repository.
npm downloads106Package downloads in the last 30 days.
Last commit1mo agoMost recent push to the default branch.
Availability100%Our rolling endpoint + install checks that succeeded.
Install checkPassedOur sandbox started it and listed its tools.
Tools exposed7Callable tools this server registers over MCP.
Directory activity1 viewsConfig copies, upvotes, and views on AllMCPs.

How MCP Server works

What declaw-ai/mcp-server MCP server does

The declaw-ai/mcp-server MCP server exposes Declaw sandbox operations through MCP. It gives an AI client a way to create an isolated execution environment, run shell commands in that environment, move files into or out of it, inspect directory contents, and clean up the sandbox when work is complete.

The available operations are create_sandbox, run_command, read_file, write_file, list_files, kill_sandbox, and list_sandboxes. Sandbox creation accepts configurable security policies. The server is intended for agent workflows that need execution isolation rather than direct access to the host system.

How it works

An MCP client starts the server with npx -y @declaw/mcp-server and supplies a Declaw API key. The client can then use the exposed tools to manage sandbox lifecycles and perform commands or file operations within a selected sandbox.

Security behavior is selected when a sandbox is created. The none preset provides no guardrails and allows full internet access. standard, the default, enables PII scanning and audit logging while retaining full internet access. strict enables PII scanning, prompt-injection defense, audit logging, and a deny-all network policy.

Outbound access can also be narrowed with allowed_domains. For example, a strict sandbox can allow selected domains such as pypi.org and github.com rather than permitting unrestricted network access.

Setup and configuration

Install the server through an MCP client configuration. Claude Desktop, Cursor, and Windsurf use a configuration entry with command set to npx, arguments -y and @declaw/mcp-server, and the API key in the environment. Claude Code can register it with claude mcp add declaw -- npx -y @declaw/mcp-server.

DECLAW_API_KEY is required. Set DECLAW_DOMAIN when connecting to a self-hosted or on-premises Declaw deployment with a custom API domain. The same MCP configuration pattern supports both hosted and on-premises use.

Tools and capabilities

  • Create sandboxes with configurable security policies.
  • Execute shell commands inside an existing sandbox.
  • Read, write, and list sandbox files.
  • List active sandboxes and destroy a sandbox.
  • Apply none, standard, or strict security presets.
  • Restrict outbound requests with an allowed-domain list.

Limitations and notes

Network behavior depends on the selected preset and any configured domain allowlist: none and standard permit full internet access, while strict starts with network access denied.

The documented interface covers sandbox management, command execution, and file operations. It does not list a separate MCP tool for snapshots, despite the README mentioning snapshots in a comparison table. Use the documented tools when determining whether the server fits a workflow.

For self-hosted deployments, configure DECLAW_DOMAIN with the custom Declaw API domain. The README identifies Apache-2.0 as the project license.

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about MCP Server

declaw-ai/mcp-server is an MCP server that connects AI tools to Declaw secure sandboxes. Its main tools create and destroy sandboxes, run shell commands, read and write files, list directory contents, and list active sandboxes. Sandbox policies can add PII scanning, prompt-injection defense, audit logging, and network restrictions.

Install it through an MCP client using `npx -y @declaw/mcp-server` and set `DECLAW_API_KEY` in the client environment. Claude Code can register it with `claude mcp add declaw -- npx -y @declaw/mcp-server`.

Yes, declaw-ai/mcp-server works with Cursor. Add an MCP configuration entry that runs `npx -y @declaw/mcp-server` and provides `DECLAW_API_KEY` in the environment.

Set a sandbox to the `strict` security preset and provide `allowed_domains` for approved outbound domains. Strict mode uses a deny-all network policy, while the allowlist can permit specific domains such as `pypi.org` or `github.com`.

Related MCP servers

Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewMCP Server AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/mcp-server-112?style=directory)](https://allmcps.com/mcp/mcp-server-112)
HTML Embed
<a href="https://allmcps.com/mcp/mcp-server-112"><img src="https://allmcps.com/api/badge/mcp-server-112?style=directory" alt="MCP Server on AllMCPs" /></a>