SkillAgentSearch skills...

declaw-cli

Command-line interface for Declaw — security-first sandboxing for AI agents

Install / Use

claude mcp add declaw-ai -- npx -y github:declaw-ai/declaw-cli

If the server publishes to npm under a different name, use that package instead — check the repo README.

About this skill
🔌

MCP Server

Model Context Protocol server

Quality Score

78/100

Category

Security

Supported Platforms

Claude Code
Claude Desktop

Our assessment of declaw-cli

declaw-cli scores 78/100 on our quality scale, 146th of 238 Security skills we index.

Its MCP Server is 6.1 KB long, well organised into 20 sections with 12 code examples: a thorough specification that gives an agent plenty to work with.

It has 3 GitHub stars, so there is little community track record yet; judge it on its content.

Substance
29/30
Structure
20/20
Description
12/15
Adoption
3/20
Freshness
15/15

Maintenance, license and trust

  • The repository was last updated yesterday, so declaw-cli is actively maintained.
  • Our last check on 2026-09-16 found the source still online.
  • It is released under the Apache-2.0 license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 87/100, with 2 cautions from licensing, adoption, age or documentation. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

No issues found

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands (1 minor note below). An AI review of the same text found nothing harmful.

  • noteInstalls by piping a downloaded script into a shellline 66
    curl -fsSL https://raw.githubusercontent.com/declaw-ai/declaw-cli/main/install.sh | sh

AI review by kimi-k2.7-code on 2026-09-25. Automated pattern scan on 2026-09-25. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

declaw-cli compared with similar skills

All 4 of these similar skills score higher than declaw-cli; compare them before choosing.

SkillScoreStarsUpdatedFormat
declaw-cli (this skill)by declaw-ai7831d agoMCP Server
Agent-Reachby Panniantong10085.3k9d agoCLAUDE.md
headroomby headroomlabs-ai10073.7ktodayCLAUDE.md
rufloby ruvnet10073.2ktodayCLAUDE.md
CowAgentby zhayujie10047.1ktodayCLAUDE.md

Frequently asked questions

How do I install declaw-cli?
Run claude mcp add declaw-ai -- npx -y github:declaw-ai/declaw-cli. The install tabs above show the steps for each supported agent.
Which AI agents does declaw-cli work with?
It is written for Claude Code and Claude Desktop, as a MCP Server file. Other agents that read the same format can often use it too.
Is declaw-cli safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands (1 minor note below). An AI review of the same text found nothing harmful. It is Apache-2.0-licensed and scores 87/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is declaw-cli still maintained?
The repository was last updated yesterday, so declaw-cli is actively maintained.

Declaw CLI

Sandbox any MCP server in one line. Drop-in for Claude Desktop, Cursor, Windsurf, Claude Code, and every MCP client.

Before / After

Your existing MCP config — no sandbox:

{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_..." }
    }
  }
}

Add declaw mcp in front — now it runs in a Firecracker microVM:

{
  "mcpServers": {
    "github": {
      "command": "declaw",
      "args": ["mcp", "--env", "GITHUB_PERSONAL_ACCESS_TOKEN", "--network-allow", "registry.npmjs.org,api.github.com,github.com,codeload.github.com", "--", "npx", "-y", "@modelcontextprotocol/server-github"],
      "env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_..." }
    }
  }
}

The MCP server runs inside a hardware-isolated sandbox. Only the environment variables you explicitly forward with --env reach the sandbox, and network is deny-all unless you allowlist specific hosts.

Why

MCP servers run as subprocesses with full host access — no sandbox, no permission model. Claude Desktop Extensions had a zero-click RCE rated CVSS 10/10 (LayerX, Feb 2026). Cursor had CVE-2025-54135 (CurXecute, CVSS 9.8) and CVE-2025-54136 (MCPoison, CVSS 8.8). declaw mcp wraps any stdio MCP server in a Firecracker microVM — the server runs unchanged, it just can't touch your machine.

Install

Option 1 — Download binary

Grab the latest from Releases. Pick the right binary for your platform:

| Platform | Binary | |----------|--------| | macOS Apple Silicon | declaw-darwin-arm64 | | macOS Intel | declaw-darwin-amd64 | | Linux x86_64 | declaw-linux-amd64 | | Linux ARM64 | declaw-linux-arm64 | | Windows x86_64 | declaw-windows-amd64.exe | | Windows ARM64 | declaw-windows-arm64.exe |

Download and move to your PATH:

chmod +x declaw-darwin-arm64
sudo mv declaw-darwin-arm64 /usr/local/bin/declaw

Option 2 — Shell script

curl -fsSL https://raw.githubusercontent.com/declaw-ai/declaw-cli/main/install.sh | sh

Option 3 — Go install

go install github.com/declaw-ai/declaw-cli/cmd/declaw@latest

Option 4 — Build from source

git clone https://github.com/declaw-ai/declaw-cli.git
cd declaw-cli && make build

After install, sign up and authenticate:

# 1. Create a free account at https://console.declaw.ai
# 2. Copy your API key from the dashboard
# 3. Authenticate:
declaw auth login

Client Setup

Claude Desktop

Config path: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows)

{
  "mcpServers": {
    "github": {
      "command": "declaw",
      "args": ["mcp", "--env", "GITHUB_PERSONAL_ACCESS_TOKEN", "--network-allow", "registry.npmjs.org,api.github.com,github.com,codeload.github.com", "--", "npx", "-y", "@modelcontextprotocol/server-github"],
      "env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_..." }
    }
  }
}

Cursor

Config path: ~/.cursor/mcp.json

{
  "mcpServers": {
    "github": {
      "command": "declaw",
      "args": ["mcp", "--env", "GITHUB_PERSONAL_ACCESS_TOKEN", "--network-allow", "registry.npmjs.org,api.github.com,github.com,codeload.github.com", "--", "npx", "-y", "@modelcontextprotocol/server-github"],
      "env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_..." }
    }
  }
}

Windsurf

Config path: ~/.codeium/windsurf/mcp_config.json — same JSON structure as above.

Claude Code

claude mcp add github -e GITHUB_PERSONAL_ACCESS_TOKEN=ghp_... -- declaw mcp --env GITHUB_PERSONAL_ACCESS_TOKEN --network-allow registry.npmjs.org,api.github.com,github.com,codeload.github.com -- npx -y @modelcontextprotocol/server-github

Flags

| Flag | Default | Description | |------|---------|-------------| | --network-allow <hosts> | deny-all | Comma-separated outbound hostname allowlist | | --template <name> | mcp-server | Sandbox template (includes Node.js + Python) | | --timeout <seconds> | 3600 | Sandbox timeout (default 1h) | | --env KEY or --env KEY=VAL | — | Environment variable to forward (repeatable). KEY reads from host env; KEY=VAL sets explicitly. | | --file LOCAL:REMOTE | — | Upload a local file into the sandbox before starting (repeatable) | | --verbose | off | Diagnostic logging to stderr |

Network is deny-all by default. MCP servers that connect to external APIs (GitHub, Slack, Brave Search, etc.) need --network-allow to reach their endpoints. This is the key security property: credentials passed to the server can only reach hosts you explicitly permit.

Custom dependencies

The default mcp-server template includes Node.js and Python, which covers most MCP servers. If your server needs additional system packages (e.g., ffmpeg, chromium, native libraries), build a custom template:

# Create a Dockerfile
echo 'FROM declaw/mcp-server:latest
RUN apt-get update && apt-get install -y ffmpeg' > Dockerfile

# Build it (returns a template ID)
declaw template build --dockerfile Dockerfile

# Use the template ID from the build output
declaw mcp --template <template-id> -- your-server-command

See declaw template build --help for details.

General Commands

declaw sandbox create|list|info|kill   Manage sandboxes
declaw sandbox exec|connect|files      Interact with sandboxes
declaw sandbox pause|resume            Lifecycle management
declaw template list|build|info|delete Manage templates
declaw volume create|list|get|delete   Manage volumes
declaw account info|usage|api-keys     Account management
declaw auth login|logout|status        Authentication
declaw mcp -- <command>                Sandbox an MCP server

Use declaw <command> --help for details on any command. All list/info commands support --json for machine-readable output.

Links

License

Apache 2.0 — see LICENSE for details.

Related Skills

View on GitHub
GitHub Stars3
CategorySecurity
Updated1d ago
Forks0

Languages

Go

Trust signals

87/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

2 low