AXI skill
Ten design principles for CLIs that agents use, and a skill that guides an agent building or reviewing such tools
- 2,241 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skill is text instructions and runs nothing and makes no network calls itself
- The reference gh-axi and chrome-devtools-axi are installed separately and give the agent access to GitHub and a controlled browser
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/axiDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill is text instructions and runs nothing and makes no network calls itself
- The reference gh-axi and chrome-devtools-axi are installed separately and give the agent access to GitHub and a controlled browser
This is third-party code. Review the repository files before installing.
What it does
AXI (Agent eXperience Interface) describes how to make a command-line tool convenient for an agent rather than for a person at a terminal. The premise is simple: every field in the output costs tokens and every extra call costs more. So the principles ask for compact TOON output, minimal fields in lists, truncated long text with a hint about the --full flag, precomputed counts instead of repeat queries and an explicit "0 results" instead of an empty reply. The axi skill puts these rules and examples into the agent's context when it writes, changes or reviews a CLI. The repository also holds a catalog of ready tools in this style, including gh-axi for GitHub and chrome-devtools-axi for the browser, plus two benchmarks with published runs.
Who it is for. For developers who build command-line tools and API wrappers for Claude Code, Codex and other agents and want to stay within a token budget.
Good fit when
- You are writing a CLI wrapper around an internal service and the agent should use it without reading a long help page
- Your tool's output bloats the context: ten-field lists, whole ticket bodies, empty replies with no explanation
- You need to audit an existing CLI against the principles: interactive prompts, unclear exit codes, unknown flags silently ignored
- You are choosing between an MCP server and a command-line tool for an agent and want published measurements of cost and turns
- You want to add the ready-made gh-axi or chrome-devtools-axi from the repository catalog to a project
Not a fit when
- You need a ready tool for a specific job: the skill runs nothing and only describes rules, the tools are installed separately from the catalog
- Your tool is for humans and agents will not call it: the token and no-interaction principles do not apply
- You are building an MCP server: the principles describe CLIs an agent runs through the shell
- You need one output format for people and machines: TOON saves tokens but is harder for ordinary scripts to consume than plain JSON
Example request
Read the axi skill and review our tasks CLI against the ten principles: show which list fields are redundant and where output should be truncatedLimitations
The skill holds recommendations and examples; it does not automate checking a tool against AXI. The published benchmarks were run on Claude Sonnet 4.6 in two areas, GitHub and browser control, and the author notes that other models may behave differently. TOON is a new format, so read its specification before building output on it; the skill asks for this in its first section. The principles target tools an agent runs through the shell, so requirements such as dropping interactive prompts cannot be carried over to ordinary human-facing programs without rework.
How to disable. Delete the axi folder from the project's skills directory (for example .claude/skills/axi), or from the global directory if you installed with -g. gh-axi and chrome-devtools-axi, installed separately, are removed with npm uninstall -g.
FAQ
What is AXI and why does an agent need it?
AXI is a set of ten principles for command-line tools that an agent runs. The goal is fewer tokens and turns: compact output, hints about the next step and structured errors with no interactive prompts.
How do I install the AXI skill in Claude Code?
Run npx skills add kunchenguid/axi. By default the skill lands in the current project's skills directory, for example .claude/skills. Add the -g flag to install it globally.
What is TOON and why does AXI require it for output?
TOON (Token-Oriented Object Notation) is an output format that the skill estimates at about 40 percent fewer tokens than equivalent JSON while staying readable for an agent. Inside a tool you can keep the logic on JSON and convert to TOON at the output boundary.
How is AXI different from MCP?
MCP is a protocol in which an agent calls tools through a server. AXI is a set of rules for ordinary CLIs that an agent runs in the shell. In the repository benchmark gh-axi finished 17 GitHub tasks at 100 percent success, while GitHub MCP reached 87 percent at nearly three times the average cost.
Which ready tools in the AXI style exist?
Official ones in the repository catalog: gh-axi for GitHub, chrome-devtools-axi for the browser, lavish-axi for reviewing HTML artifacts and quota-axi. The community added, for example, npm-axi for the npm registry, homebrew-axi for Homebrew, playwright-axi and bun-axi.
Discussion
No questions or reviews yet.
Related
SkillsA CLI installer and package manager for coding agent skills: one command installs skills into many runtimes and helps build your own
Axe in Go: single-purpose AI agents defined in TOML and run from a terminal, pipes, cron and git hooks, with memory, sub-agents and MCP
A Go CLI that configures the AI coding agents you already use: it gives them shared memory, a workflow and deterministic steps without replacing the agents them
Agents Best Practices
agents-best-practices
A provider-neutral skill for designing agent systems: loop architecture, tools, permissions, memory and observability
A CLI that keeps coding agents' shared context in one place: instructions, skills, MCPs, subagents and durable memory across projects and machines
A multi-agent harness that verifies results by artifacts: stop hooks, an independent judge and an event log for Claude Code, Codex, Cursor and more
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| AXI (this entry) | no | 2,241 | ||
| Context Mode | no | not required | 31,493 | compresses tool output during a session, while AXI describes how to build the CLI itself to be economical |
| Chrome DevTools MCP | yes | not required | 53,919 | an MCP server for browser control, while AXI has the chrome-devtools-axi wrapper for the same task |
README in short
The repository opens with two benchmarks on Claude Sonnet 4.6: 14 browser tasks and 17 GitHub tasks, five repeats each, with a table of success, cost, time and turns. A quick start follows with the installation of gh-axi and chrome-devtools-axi and a line for CLAUDE.md or AGENTS.md. Then come the ten principles and a catalog of tools, official and community, with a domain and short description. The "Build Your Own AXI" section points to the skill and recommends an explicit session-hook setup command as the primary integration, with the skill as a lighter second path. The end lists commands to run both benchmarks.
README badge
Original SKILL.md
---
name: axi
description: >
Agent eXperience Interface (AXI) — ergonomic standards for building CLI tools that agents
use via shell execution. Use when building, modifying, or reviewing any agent-facing CLI.
---
# Agent eXperience Interface (AXI)
AXI defines ergonomic standards for building CLI tools that autonomous agents interact with through shell execution.
## Before you start
Read the [TOON specification](https://toonformat.dev/reference/spec.html) before building any AXI output.
## 1. Token-efficient output
Use [TOON](https://toonformat.dev/) (Token-Oriented Object Notation) as the output format on stdout.
TOON provides ~40% token savings over equivalent JSON while remaining readable by agents.
Convert to TOON at the output boundary — keep internal logic on JSON.
```
tasks[2]{id,title,status,assignee}:
"1",Fix auth bug,open,alice
"2",Add pagination,closed,bob
```
## 2. Minimal default schemas
Every field in stdout costs tokens — multiplied by row count in collections.
Default to the smallest schema that lets the agent decide what to do next: typically an identifier, a title, and a status.
- Default list schemas: 3-4 fields, not 10
- Default limits: high enough to cover common cases in one call (if most repos have <100 labels, default to 100, not 30)
- Long-form content (bodies, descriptions) belongs in detail views, not lists
- Offer a `--fields` flag to let agents request additional fields explicitly
## 3. Content truncation
Detail views often contain large text fields. Omitting them forces agents to hunt; including them wastes tokens.
Truncate by default and tell the agent how to get the full version.
```
task:
number: 42
title: Fix auth bug
state: open
body: First 500 chars of the issue body...
... (truncated, 8432 chars total)
help[1]: Run `tasks view 42 --full` to see complete body