I built pmctl, a CLI for browsing Postman collections from the terminal. This week I turned it into something more interesting: a skill that AI coding agents can use to discover and call your APIs.

The problem

Every team I’ve worked with has hundreds of Postman requests sitting in shared workspaces. When you’re pair-programming with an AI agent – Claude Code, Cursor, Codex – and you need to hit an API, you end up copy-pasting URLs, headers, and auth tokens from Postman into your conversation. It’s tedious.

What if the agent could just browse your Postman workspace itself?

What pmctl does

It’s a Python CLI (pip install pmctl) that wraps the Postman API:

# Browse collections
pmctl collections list

# Fuzzy search for requests
pmctl requests list -c "My API" --search "getUser"

# Get full request details (headers, body, params)
pmctl requests show "create User" -c "My API"

# Resolve environment variables to real URLs
pmctl environments show "Production" --json

The --json flag makes everything machine-readable, which is exactly what AI agents need.

Making it agent-native

The CLI alone isn’t enough. AI agents need instructions – they need to know when to use a tool and how to use it effectively. That’s what the Agent Skills spec is for: a SKILL.md file that agents load when relevant.

I wrote one for pmctl and published it everywhere I could find:

  • ClawHub – OpenClaw’s skill registry
  • Context7 – skills registry with trust scoring (pmctl got a 9.4)
  • Cursor Marketplace – submitted, pending review
  • Claude Code – installable as a plugin marketplace
  • OpenAI Codex – .agents/skills/ directory convention

The skill teaches agents workflows like “resolve a Postman \{\{base-url\}\} variable for a specific environment” or “construct a curl command from a Postman request.” These are the patterns developers do manually dozens of times a day.

The agent skills ecosystem is exploding

While publishing pmctl, I mapped out the current landscape. It’s early 2026 and there are already:

Most of these registries auto-index from GitHub. If your repo has a skills/*/SKILL.md file following the Agent Skills spec, crawlers will find it.

The interesting part: the same SKILL.md works across Claude Code, Codex, Cursor, Gemini CLI, and OpenCode. Write once, run everywhere. The tooling differences are just about where you put the file:

Tool Skill location
Claude Code skills/*/SKILL.md + .claude-plugin/
Cursor .cursor/rules/*.mdc or skills/*/SKILL.md + .cursor-plugin/
Codex .agents/skills/*/SKILL.md
OpenClaw skill/SKILL.md

Try it

pip install pmctl
pmctl profile add mywork --api-key "PMAK-..." --default

Or install the agent skill directly:

# Context7
npx ctx7 skills install /wbingli/pmctl pmctl

# Claude Code
claude plugin marketplace add wbingli/pmctl
claude plugin install pmctl@pmctl

# ClawHub
clawhub install pmctl

The repo is at github.com/wbingli/pmctl. PRs welcome.