AI Native Lang
Back to Learn
general

How to install & set up AINL with ZeroClaw

Install ZeroClaw first (Homebrew, install.sh, or from source), then add the AINL skill or ainl install-mcp --host zeroclaw for MCP and ainl-run.

March 21, 2025·4 min read
#zeroclaw#install#mcp
Share:TwitterLinkedIn

How to install & set up AINL with ZeroClaw

This guide is in two parts: (1) get ZeroClaw running on your machine, (2) wire in AINL (skill + MCP + CLI). If you skip part 1, commands like zeroclaw skills install will not exist yet.

Not sure which stack you use? Start from Easy install and pick OpenClaw or ZeroClaw there.


Part 1 — Install ZeroClaw (do this first)

ZeroClaw is a separate project from AINL. Use the official repository only:

Impersonation warning: Other domains or repos claiming to be “ZeroClaw” may be unrelated. If something does not match the zeroclaw-labs/zeroclaw repo or zeroclawlabs.ai, do not trust binaries or install scripts from it.

What you need before AINL

  • ZeroClaw CLI on your PATH (the zeroclaw command works in a new terminal).
  • A provider / API key configured (ZeroClaw onboarding walks you through this). AINL does not replace that—you still run the assistant through ZeroClaw.

Step 1 — Pick an install method

A) macOS / Linux: Homebrew (simplest if you use Brew)

brew install zeroclaw

B) One-liner installer (downloads upstream install.sh)

curl -LsSf https://raw.githubusercontent.com/zeroclaw-labs/zeroclaw/master/install.sh | bash

C) Git clone + install script

git clone https://github.com/zeroclaw-labs/zeroclaw.git
cd zeroclaw
./install.sh

After ./install.sh, many setups run zeroclaw onboard automatically. If yours does not, run it yourself (recommended official path):

zeroclaw onboard

D) Windows

ZeroClaw’s docs expect Windows via WSL2 (same Linux flow as above) plus Rust / Visual Studio Build Tools when building from source. See the Windows prerequisites section in the official README; prefer prebuilt or install.sh inside WSL if you are not developing ZeroClaw itself.

Step 2 — Quick verification

Open a new terminal and check the CLI:

zeroclaw --version
zeroclaw status
# optional health check:
zeroclaw doctor

Optional smoke test (after onboarding / API key):

zeroclaw gateway   # control plane + dashboard (default bind is documented in ZeroClaw docs)
zeroclaw agent -m "Hello from ZeroClaw"

If zeroclaw is not found, finish ZeroClaw’s install docs first—do not continue to Part 2 until the binary is on your PATH.

Step 3 — Where to read more


Part 2 — Install AINL on top of ZeroClaw

AINL integrates via a skill (Markdown importer, runtime shim, ainl-mcp) and ainl install-mcp --host zeroclaw, which merges ~/.zeroclaw/mcp.json, installs ~/.zeroclaw/bin/ainl-run, and updates your shell PATH when possible.

Prerequisites (AINL)

  • Python 3.10+ and pip
  • Working zeroclaw CLI (Part 1)
  • Network access to PyPI and (for the skill URL) GitHub

Path A — ZeroClaw skill from the main repo

Install the skill from the published tree (recommended):

zeroclaw skills install https://github.com/sbhooley/ainativelang/tree/main/skills/ainl

Then run the skill’s installer (pulls / upgrades ainl[mcp] and wires local expectations):

cd <skill-checkout> && ./install.sh

From a full clone of the monorepo:

cd skills/ainl && ./install.sh

When the skill is published as a standalone repo, you can use:

zeroclaw skills install https://github.com/sbhooley/ainl-zeroclaw-skill
cd <skill-checkout> && ./install.sh

Path B — Bootstrap without the skill checkout

Install the package and run the ZeroClaw-oriented bootstrap:

pip install 'ainativelang[mcp]'
ainl install-mcp --host zeroclaw
  • Preview changes: ainl install-mcp --host zeroclaw --dry-run (same as install-zeroclaw --dry-run)
  • Verbose logs: ainl install-mcp --host zeroclaw --verbose

This typically:

| Artifact | Purpose | | --- | --- | | pip install --upgrade 'ainativelang[mcp]' | Compiler, importer extras, MCP deps | | ~/.zeroclaw/mcp.json | Merges an ainl stdio server entry pointing at ainl-mcp (skipped if an equivalent entry exists) | | ~/.zeroclaw/bin/ainl-run | Wrapper: ainl compile "$1" && ainl run "$1" (extra args forward to ainl run) | | ~/.bashrc / ~/.zshrc | Appends export PATH="$HOME/.zeroclaw/bin:$PATH" when those files exist and don’t already mention ~/.zeroclaw/bin |

If no rc file is updated, the command prints a one-line PATH tip you can paste manually.

Ensure ainl-mcp is on your PATH (same venv as ainl), or use the full path in mcp.json.

Chat example

After a successful install, try:

Import the morning briefing using AINL.

Then use ainl import markdown …, ecosystem shortcuts (ainl import clawflows / ainl import agency-agents), or MCP tools (ainl_list_ecosystem, ainl_import_clawflow, ainl_import_agency_agent, ainl_import_markdown) so the agent produces compiling .ainl source, followed by ainl compile / ainl run or:

~/.zeroclaw/bin/ainl-run my.ainl

(with ~/.zeroclaw/bin on PATH, you can run ainl-run my.ainl).

Docs & source

If you use OpenClaw instead, follow How to install & set up AINL with OpenClaw.

A

AI Native Lang Team

The team behind AI Native Lang — building deterministic AI workflow infrastructure.

Related Articles