Skip to content

setup-mcp-install.md

Latest commit

 

History

History
36 lines (24 loc) · 1.72 KB

setup-mcp-install.md

File metadata and controls

36 lines (24 loc) · 1.72 KB
Raw

safedep setup mcp install

Guided first-time onboarding: authenticate with SafeDep Cloud and inject the SafeDep MCP server into AI coding agent config files in a single command.

Agents supported: Claude Code, Cursor, Gemini CLI, OpenCode, Antigravity.

Synopsis

safedep setup mcp install [flags]

Flags

Flag Description
--mcp-url <url> SafeDep MCP server URL (default: https://mcp.safedep.io/model-context-protocol/threats/v1).
--workspace <dir> Project directory for workspace-level injection. Empty (default) skips workspace injection.
--force Bypass credential check and always re-authenticate via device flow.

Inherits root flags --output, --profile, and --insecure-keychain-fallback.

What it does

  1. Checks for existing credentials (API key + tenant) in the active profile's keychain. If found and --force is not set, skips authentication and goes directly to agent configuration.
  2. If no credentials exist (or --force is set): runs the OAuth2 device-code flow, prompts for tenant selection (and registration if the account is new), creates an API key, and saves everything to the keychain.
  3. Derives the machine's stable endpoint identity (X-Endpoint-ID) from the hardware UUID and hostname.
  4. Detects which supported AI agents are installed on this machine.
  5. Writes the mcpServers.safedep entry into each detected agent's config file. The write is idempotent.

If authentication succeeds but agent configuration fails, credentials are kept and an advisory message is printed. Use safedep protect mcp install to retry the configuration step.

Exit codes

  • 0 on success, including the case where no agents are detected.
  • 1 on authentication or keychain error.