Agent skill
Install the SKILL.md so your agent knows how to query the directory, read the two trust axes, and complete payment handoff per rail.
Install
# Claude mkdir -p ~/.claude/skills/agentic-terminal curl -fsSL https://agenticterminal.ai/SKILL.md \ -o ~/.claude/skills/agentic-terminal/SKILL.md
Connect — three paths
# A. Hosted MCP (recommended, no install)
{ "mcpServers": { "at-directory": { "url": "https://mcp.agenticterminal.ai/mcp" } } }
# B. Local MCP (npm, Node >= 20)
npm install -g @agenticterminal/mcp-server
# C. Plain REST — for agents that don't speak MCP
curl 'https://mcp.agenticterminal.ai/v1/merchants?rail=lightning'
curl 'https://mcp.agenticterminal.ai/v1/merchants/bitrefill'Reads are anonymous and ungated on all three — no credential needed to discover or transact. REST endpoints return the same JSON as the matching MCP tool.
The two axes, restated
OP trust tier describes the merchant. Agent-callable tier describes the integration. They are independent — always read both before transacting.
Full skill text: agenticterminal.ai/SKILL.md(served from the repo's packages/skill/SKILL.md).