514 agent init

Install agent skills, configure MCP servers, set up the MooseStack LSP, and install agent-specific plugins for your coding agents. Supports 21+ agents including Claude Code, Cursor, VS Code, Windsurf, Codex, and more.

Usage

514 agent init [OPTIONS]
514 agent init schema [--json]

What gets installed

When you run 514 agent init, the following components are configured:

• Skills: Downloaded from github.com/514-labs/agent-skills and stored at ~/.agents/skills/. Symlinks are created in each agent's skills directory (e.g. ~/.claude/skills/, ~/.cursor/skills/).
• MooseStack LSP: The MooseStack language server is installed globally via npm, providing SQL intelligence in TypeScript projects.
• MCP servers: Configuration for the MooseDev MCP server (localhost:4000) and Context7 (mcp.context7.com) is written to each agent's MCP config file.
• Plugins: Agent-specific plugins are installed where supported (e.g. Claude Code marketplace plugin).

Interactive mode

By default, 514 agent init runs interactively. It auto-detects installed coding agents by checking for known configuration directories, fetches available skills from GitHub, and prompts for confirmation before installing.

# Auto-detect agents and install everything
514 agent init
 
# Target specific agents instead of auto-detecting
514 agent init --agent claude-code --agent cursor
 
# Install skills from a specific branch
514 agent init --branch feature/new-skills

Non-interactive mode

Non-interactive mode is activated when --json is passed or when stdin is not a terminal (piped input). In non-interactive mode, explicit confirmation is required via the --yes flag or "yes": true in the JSON input.

Input transports

There are two ways to provide structured JSON input:

1. File: Pass --input <path> to read JSON from a file.
2. Stdin: Pipe JSON to stdin when no --input flag is provided.

CLI flags always take precedence over JSON fields.

Input JSON format

{  "version": 1,  "agents": ["vscode", "claude-code"],  "branch": "main",  "yes": true}

Examples

# Non-interactive with flags
514 agent init --yes --json
 
# Target a specific agent with flags
514 agent init --agent vscode --yes --json
 
# Read input from a file
514 agent init --input request.json --yes --json
 
# Pipe JSON via stdin
echo '{"version":1,"agents":["vscode"],"yes":true}' | 514 agent init --json

JSON output

On success:

{  "status": "ok",  "installed": [    {      "skill": "514/cli",      "agents": [        {          "id": "vscode",          "name": "VS Code",          "path": "~/.vscode-copilot/skills/514--cli"        }      ],      "canonical_path": "~/.agents/skills/514/cli"    }  ],  "lsp_installed": true}

On error:

{  "status": "error",  "message": "514 agent init requires explicit confirmation in non-interactive mode; pass --yes or set \"yes\": true in the input JSON"}

Schema discovery

Use the schema subcommand to discover available input options programmatically. This is useful for IDE extensions and automation tools that need to call 514 agent init without hardcoding the input contract.

# Human-readable summary
514 agent init schema
 
# Machine-readable JSON schema
514 agent init schema --json

The JSON schema includes supported agent IDs, field definitions, transport options, and example inputs. Extensions can call this command to dynamically discover what options are available, so they don't need code changes when new options are added.

Options

• --api-url <API_URL>: API base URL override (HOSTING_CLI_API_URL).
• -o, --org <ORG>: Organization name or ID.
• --json: Output as JSON.
• -v, --verbose...: Enable verbose output (-v debug, -vv trace).
• -h, --help: Print help.

Supported agents

All agents receive skills via symlinks. MCP indicates that MCP server configuration is written to the agent's config file. LSP indicates that language server configuration is written. Plugin indicates that an agent-specific plugin is installed.

Related

• 514 agent
• 514 agent remove
• MooseStack LSP
• MooseDev MCP