---
title: Skill — Install & Use
description: Drop the OpenMobius-skill into Claude Code / Codex / OpenClaw / Hermes — one command, agent gains MobiusQuant superpowers.
---

# Skill Integration

> The fastest way to put MobiusQuant inside your AI agent. One install command, four platforms supported, your agent becomes a market-aware quant in <10 minutes.

[![Repo](https://img.shields.io/badge/GitHub-OpenMobius--skill-171717?logo=github)](https://github.com/MobiusQuant/OpenMobius-skill)
[![License](https://img.shields.io/badge/license-Apache_2.0-blue.svg)](https://github.com/MobiusQuant/OpenMobius-skill/blob/main/LICENSE)
[![Python](https://img.shields.io/badge/python-3.10%2B-3776ab.svg)](https://www.python.org/)

## What you get

See it in action — one natural-language prompt drives install + first analysis end-to-end:

![OpenMobius-skill end-to-end demo](/demo/final-combined.gif)

After install, the AI agent in your terminal can:

| Ask it... | And it... |
| --- | --- |
| *"What is Fair Value Gap, how to trade it?"* | Vector-retrieves FVG concept card + related (CISD / OTE / Premium-Discount), cites rules from the knowledge base |
| Attach a BTCUSDT 1h chart + *"analyze this"* | Fetches real OHLCV, extracts FVG / OB / sweep / displacement, replies in 5 sections with **exact prices** + auto-annotated PNG |
| *"How is BTC 1h looking?"* | Live data fetch → KB-grounded analysis |
| *"What's RSI(14) and MACD on BTC?"* | Calls indicator endpoint + applies per-indicator analysis dimensions |
| Paste OHLCV CSV | Parses → analyzes → KB cross-reference → 5-section reply |
| *"Generate a chart with my entry/SL/target"* | Renders via Playwright + lightweight-charts |

Backed by:
- **964 curated knowledge cards** (380 concepts + 584 cases distilled from 130 ICT/SMC teaching videos)
- **Real-time market data** (Binance · Bybit · OKX · Hyperliquid · A-shares · HK · US · forex)
- **60+ technical indicators** with built-in interpretation rubric (`summary_focus`)

## Install

### Recommended — let your AI agent install it (one prompt)

Open the AI agent you already use (Claude Code / Codex / OpenClaw / Hermes) and paste this:

```
Install OpenMobius-skill on my machine. Follow this playbook exactly:
https://github.com/MobiusQuant/OpenMobius-skill/blob/main/README_AGENT.md
```

The agent runs the 6-step playbook automatically: **preflight check → platform detection → warn about download size → run installer → health check → report back**. It pauses for confirmation when needed. Takes 5–10 minutes (mostly model + dependency downloads).

::: tip Other phrasings work too
The agent recognises a range of natural-language triggers — any of these will start the install:

- "install OpenMobius"
- "install the mobius trading skill"
- "set up OpenMobius-skill on this machine"
- "帮我装一下 OpenMobius"

We wrote [`README_AGENT.md`](https://github.com/MobiusQuant/OpenMobius-skill/blob/main/README_AGENT.md) specifically for AI agents — it's a step-by-step playbook covering the per-platform path table, success criteria for every step, failure handling, and post-install verification. **You don't need to touch the command line.**
:::

### Advanced — run the commands yourself

If you'd rather drive the installer manually:

```bash
git clone https://github.com/MobiusQuant/OpenMobius-skill.git
cd OpenMobius-skill
python install.py
```

The installer:

1. Creates `.venv/` and installs dependencies
2. Downloads Playwright chromium (~280 MB)
3. Downloads `nomic-embed-text-v1.5` embedding model (~274 MB)
4. Builds the 964-card vector index
5. Registers the skill into `~/.claude/skills/OpenMobius-skill/`
6. Runs a health check

**First run**: ~5–10 min · **Subsequent updates**: <30 s.

::: tip Prerequisites
Python 3.10+. See the repo's [INSTALL.md](https://github.com/MobiusQuant/OpenMobius-skill/blob/main/INSTALL.md) for details.
:::

## Platform support

```bash
python install.py --platform <name>
```

| Platform | Flag | Default install path |
| --- | --- | --- |
| **Claude Code** | `--platform claude-code` *(default)* | `~/.claude/skills/OpenMobius-skill/` |
| **Codex** | `--platform codex` | `~/.codex/skills/OpenMobius-skill/` |
| **OpenClaw** | `--platform openclaw` | `~/.openclaw/skills/OpenMobius-skill/` |
| **Hermes** | `--platform hermes` | `~/.hermes/skills/market-data/OpenMobius-skill/` |
| Auto-detect | `--platform auto` | scans `~/.<agent>` dirs |
| All four | `--platform all` | loops through all |

Shared resources (`scripts/`, `knowledge_base/`, `.venv/`, model cache) are **symlinked** — installing on 4 platforms adds only ~17 KB extra per platform.

## After install — start using

In your AI agent (Claude Code / Codex / etc.) just ask plainly:

```
"What is Liquidity Sweep"
[attach chart] "analyze this setup"
"How is ETH 4h looking, give me a chart"
"BTC 1h RSI(14) and MACD?"
```

The skill auto-invokes on description match — no manual mode-switch needed.

## API Token (optional but recommended)

Live data fetches go to `api.mobiusquant.ai` with **two tiers**:

| Mode | Rate limit | When |
| --- | --- | --- |
| Anonymous | 10 req/min per IP | Quick try, no setup |
| With token | 60 req/min | Sustained use |

Configure once and forget — the agent reads `.env.local`:

```bash
# Easiest: paste the token into the chat — the agent runs the validation +
# write + verify pipeline automatically:
You:   <paste mq_xxx... token>
Agent: ✓ Token configured (mq_xxxx…yyyy)

# Manual:
cd <install-dir>
.venv/bin/python scripts/kb_klines.py token set <TOKEN>
```

Get a free token at [/apply-token](https://www.mobiusquant.ai/zh/apply-token) (7-day anonymous) or [/account](https://www.mobiusquant.ai/zh/account) (logged-in permanent). See [Apply for a Token](/token) for full details.

## Update

```bash
# Pull latest scripts + dependencies
python install.py --update

# Update AND rebuild the vector index (run after KB content changes)
python install.py --update --rebuild-index
```

Updates are fast (<30s) because deps / models / index are reused.

## Uninstall

| Goal | Command |
| --- | --- |
| Remove platform registration only (soft, keeps `.venv` + index) | `python install.py --uninstall` |
| All platforms at once | `python install.py --uninstall --platform all` |
| Full uninstall (`.venv` + index too) | `python install.py --uninstall --full` |
| Full purge (also delete shared chromium + nomic caches — may be used by other projects) | `python install.py --uninstall --purge --yes-i-know` |

## Troubleshooting

Run the doctor first:

```bash
.venv/bin/python scripts/kb_doctor.py
```

It reports: venv state · deps · nomic model · vector index · CJK fonts · skill registration · API connectivity.

Common issues:

| Symptom | Fix |
| --- | --- |
| Chinese labels render as boxes | Install `fonts-noto-cjk` (Linux); macOS / Windows usually bundled |
| API request fails | Check network; probe `https://api.mobiusquant.ai/api/health` |
| Skill not auto-invoking in Claude Code | Verify `~/.claude/skills/OpenMobius-skill/` exists; restart the agent |
| `chroma.sqlite3` not found | Re-run `.venv/bin/python scripts/build_index.py` |

## How it works (high-level)

```
OpenMobius-skill/
├── SKILL.md                # main entry — your AI agent reads this
├── SKILL.body.md           # shared body (platform-neutral)
├── platforms/              # per-platform frontmatter (claude-code / codex / openclaw / hermes)
├── workflows/              # qna / analyze / annotate / klines
├── scripts/                # CLI tools called by the AI
├── knowledge_base/         # 380 concepts + 584 cases
└── install.py              # cross-platform installer
```

The agent invokes via `SKILL.md` description match → routes to one of four workflows:

- **Q&A** — concept lookup (`kb_retrieve.py`)
- **analyze** — chart + analysis with cited rules
- **annotate** — overlay entry/SL/target onto user's chart
- **klines** — fetch + indicator + KB-grounded analysis

## Source

- Repo: <https://github.com/MobiusQuant/OpenMobius-skill>
- License: Apache 2.0
- Issues / PRs: <https://github.com/MobiusQuant/OpenMobius-skill/issues>

## Related

- [Apply for an API Token](/token) — anonymous 7-day trial or permanent (login)
- [Supported Markets](/markets) — 11 venues × 512 symbols
- [Indicator Catalog](/indicators/) — 56 indicators with knowledge-attached responses
- [AI Agent Operating Manual](/agents) — site structure for AI scraping