Installation¶
Prerequisites¶
- Claude Code CLI, installed and authenticated
- Python 3.10+, with
piporuvavailable on PATH
Option 1: npm¶
This does three things:
- Installs the
claude-turingCLI globally - Registers the plugin with Claude Code (writes to
~/.claude/plugins.json) - Runs verification to confirm all commands, agents, and Python dependencies resolve
Requires Node.js v18+.
Option 2: Local path¶
If you are developing Turing or prefer a local install:
This registers the plugin directly from a local checkout. Changes to command files take effect immediately; no reinstall needed.
Verification¶
Run the verifier to confirm everything works:
Expected output:
Turing ML Research Harness v4.7.0
==================================
Commands .................. 74/74 registered
Agents .................... 2/2 available
@ml-researcher ......... Read/Write/Edit/Bash
@ml-evaluator .......... Read/Bash (read-only)
Python dependencies ....... OK
python .................. 3.12.3
xgboost ................ 2.1.1
scikit-learn ........... 1.5.2
pandas ................. 2.2.3
Templates ................. 8/8 present
Plugin registration ....... OK
All checks passed.
If any check fails, the verifier prints the specific remediation step. Common issues:
| Symptom | Fix |
|---|---|
Python dependencies ... FAIL |
Run pip install -r requirements.txt or uv pip install -r requirements.txt in the Turing directory |
Plugin registration ... FAIL |
Run claude-turing install --global again |
Commands ... N/74 registered |
Update to latest version: npm update -g claude-turing |