Use Skillhawk where you already work
Everything here is free and open source and runs on your machine: no account, no key, nothing uploaded. You need Node.js 20 or newer.
1. Run the free check
npx skillhawk agents
Lists agents that run without you and which account each scheduled one bills: your subscription or a pay-per-token API key. It prints variable names only, never key values. More commands on the tools page.
2. Add the MCP server to your agent
Your agent gets four tools: review_skill, secure_scan, compile_skill and check_agents. They run locally over stdio. connect writes the config for you; without --write it prints it so you can check it first.
Claude Code
npx skillhawk connect claude --writeOr add it by hand
claude mcp add --scope user skillhawk -- npx -y skillhawk mcpCodex
npx skillhawk connect codex --writeOr add it by hand
# ~/.codex/config.toml
[mcp_servers.skillhawk]
command = "npx"
args = ["-y", "skillhawk", "mcp"]Cursor
npx skillhawk connect cursor --writeOr add it by hand
// .cursor/mcp.json
{
"mcpServers": {
"skillhawk": {
"command": "npx",
"args": [
"-y",
"skillhawk",
"mcp"
]
}
}
}VS Code (Copilot)
npx skillhawk connect vscode --writeOr add it by hand
// .vscode/mcp.json
{
"servers": {
"skillhawk": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"skillhawk",
"mcp"
]
}
}
}Gemini CLI
npx skillhawk connect gemini --writeOr add it by hand
// ~/.gemini/settings.json
{
"mcpServers": {
"skillhawk": {
"command": "npx",
"args": [
"-y",
"skillhawk",
"mcp"
]
}
}
}Claude Desktop
npx skillhawk connect claude-desktop --writeOr add it by hand
// claude_desktop_config.json
{
"mcpServers": {
"skillhawk": {
"command": "npx",
"args": [
"-y",
"skillhawk",
"mcp"
]
}
}
}Any other MCP client: start the server with npx -y skillhawk mcp.
3. Find Skillhawk in directories
- npm: The package: npx skillhawk, no install needed.
- Official MCP Registry: Listed as io.github.agentwik-labs/skillhawk (registry entry, JSON).
- Glama MCP directory: MCP server listing with a build test of the tools.
- GitHub: Source, issues and releases (MIT).
For agents reading this site
- llms.txt: Site summary for language models.
- Agent skill: A SKILL.md your agent can load to use Skillhawk (discovery index at /.well-known/agent-skills/index.json).
The home page also offers browser agents two read-only WebMCP tools, where the browser supports WebMCP.
Questions
Email help@skillhawk.ai or open an issue on GitHub.