Skip to main content
bounty-cli gives operators and AI coding agents a terminal-safe way to inspect Bounty data without direct database, warehouse, or infrastructure access. The CLI is one public surface of Bounty. It is open source and is especially intended for Claude Code, Codex, and other AI coding agents that need a narrow, authenticated interface for marketing workflows. For OAuth-aware MCP clients, Bounty also exposes an app-hosted MCP server with read-only tools and dynamic client registration for allowed public PKCE clients, including local loopback clients and supported hosted ChatGPT and Claude connector callbacks.

​
What The CLI Covers

  • Campaign, ad, and creative performance inspection.
  • One-campaign diagnosis with prioritized findings and evidence.
  • Generated action review, including ICE scores, owners, statuses, and verdicts.
  • Agent inventory and configuration inspection.
  • Skill installation for AI coding agents that should use the CLI consistently.

​
Agent Skills

Bounty agent skills are workflow instructions for AI coding agents. They help Claude Code, Codex, and other agents pick the right bounty-cli command, use explicit date ranges, prefer JSON output, and avoid asking for private infrastructure credentials. Install the CLI skills and authenticate in one step: 
npx -y bounty-cli@latest init --all --browser
Then restart your AI coding agent so it can discover the installed skills.

​
Principles

  • Use the logged-in user’s Bounty permissions.
  • Prefer machine-readable --json output for agent workflows.
  • Use explicit date ranges for performance claims.
  • Keep private infrastructure credentials out of agent context.

​
Next Steps