CLI

Agent mode & skills

JSON output for agent loops, and packaged skills like agentic-purchase.

The Ramp CLI is useful to humans, but it is also designed to be driven by agents. Two features matter most for agent loops: structured output and skills.

Agent mode (JSON output)

The CLI has two output modes. In a terminal it prints human-readable tables; when piped, it switches to JSON automatically. You can force either:

Flag	Behavior
`--human`	Human-readable table output (default in a terminal)
`--agent`	Machine-readable JSON output (default when piped)
`--wide`	Show all columns in table output

For non-interactive environments, combine --no-input (disable prompts) with --quiet (suppress progress output):

ramp transactions list --agent --no-input --quiet

Skills are structured instructions that teach an agent how to use Ramp for a specific workflow. They are especially useful when the agent needs to coordinate multiple Ramp calls with browser work or another app. Browse and install them with the skills command:

ramp skills list
ramp skills install agentic-purchase

The most common skill for AI-driven flows is agentic-purchase, which drives the Agent Cards capability end-to-end: request a credential, check out, and close the loop. You can also point an agent straight at the skill definition:

https://github.com/ramp-public/ramp-cli/blob/main/src/ramp_cli/skills/agentic-purchase/SKILL.md

Skills may call Ramp commands, MCP tools, or browser actions depending on the workflow. The user's Ramp permissions and scopes still decide what succeeds.

Ready-made playbooks

For copy-and-paste prompts that wrap these skills per runtime, see Playbooks.