MobiusQuant
Docs

MobiusQuant

Give your Quant AgentKnowledgeMarket DataDecisionsMemory
Infrastructure for AI trading
Make your AI agent actually understand the market, reason about trades, and remember what worked.
We package every primitive your AI trader needs into a single platform — market data, in-house indicators, an indicator knowledge base, analysis Playbooks, multi-platform Skills, and long-term memory — natively wired into Claude Code, Codex, OpenClaw, and Hermes. You focus on shaping your own AI agent. The plumbing is already done.
Telegram
Discord
X
Medium

Native integrations with the AI Agent tools you already use

Don't change your workflow. MobiusQuant ships as a Skill that drops directly into the AI Agent tools you're already using — one capability, every endpoint. All integrations are open source under Apache-2.0; PRs and issues welcome.
✨ Now open-source
MobiusQuant/OpenMobius-skill
· Apache-2.0
GitHub stars
View repo
Read README
💬 Recommended
Agent-driven install
Paste the prompt below into the AI agent you already use (Claude Code / Codex / OpenClaw / Hermes). It reads our dedicated README_AGENT.md and runs the 6-step install for you, reporting back when done.
Install OpenMobius-skill on my machine. Follow this playbook exactly:
https://github.com/MobiusQuant/OpenMobius-skill/blob/main/README_AGENT.md
The agent will: ① check Python/Git → ② detect the target platform → ③ download deps and models → ④ register the skill → ⑤ run a health check → ⑥ report back. ~5–10 minutes total; it pauses for confirmation when needed.
Claude CodeAnthropic's official CLI
Call OpenMobius-skill from inside Claude Code and pull market data, indicators, and Playbook analyses straight to your terminal — your AI trader lives in your shell.
Live
GitHub
CodexOpenAI CLI agent
Wire the same skill into Codex and run analyses inside the OpenAI workflow you already know. One capability, every endpoint.
Live
GitHub
OpenClawOpen-source agent framework
Brings the same capability to the open-source agent ecosystem. Plug MobiusQuant in as a tool node in your agent graph.
Live
GitHub
HermesNous Research agent
Available as a skill in hermes-agent.nousresearch.com. Give your Hermes agent trading-grade market perception and reasoning tools.
Live
GitHub
Or run the commands yourself
Claude Code
git clone https://github.com/MobiusQuant/OpenMobius-skill.git
cd OpenMobius-skill
python install.py
# default platform is claude-code; equivalent to:
# python install.py --platform claude-code

The hard part of AI trading isn't the AI — it's whether the AI can read the market

Builders of AI traders spend 90% of their time on plumbing — wiring market data, writing indicators, curating knowledge bases, tuning prompts, tracking strategy history. The actual «let the AI think» part barely gets a window. We finish that plumbing for you, leaving the AI agent three jobs: understand, reason, remember.

Perception layer
Eyes on the market for your AI

Candles alone can't express trend, momentum, volatility, or support/resistance. And LLMs are bad at indicator math, and worse at remembering interpretive rules — they make stuff up (Is RSI overbought at 70 or 80? How do I read this signal?).
We ship market data + indicators + an indicator knowledge base as one bundle. Each response carries not just the numbers but what to look at, how to read signals, and what pitfalls to avoid. The LLM writes the analysis straight from the rubric — no guessing, no hallucination.

Reasoning layer
Reusable mental models for your AI

«Golden cross», «oversold», «breakout» — the same signal means different things in different regimes. Forcing the AI to reinvent its reasoning every time is slow and unstable.
We provide analysis Playbook templates and multi-platform Skill extensions, distilling proven trading logic into composable, callable reasoning primitives.

Memory layer
An AI that grows into you

Every conversation, post-mortem, and strategy tweak deposits into your own trading memory.
Once that memory is thick enough, it becomes your trading mentor, your post-mortem partner, your 24/7 market watcher — your AI trading double.

Numbers + knowledge in the same response

Call `/api/indicators?explain=true` and we don't just compute the indicator — we pack the interpretation rubric right into the response. The LLM stops guessing thresholds from training memory; hallucination drops noticeably.
✕ Typical indicator API
# Typical indicator API
GET /indicators?name=cvd
{
  "data": {
    "cvd_hist":  [-12, -8, -3, 5, 11, ...],
    "bear_div":  [false, false, true, ...]
  }
}

# LLM is left guessing:
#   "What does bear_div=True mean?"
#   "How is divergence strength judged?"
#   "Is RSI overbought at 70 or 80?"
#   → invents rules from training memory,
#     hallucination risk high, output unstable
✓ MobiusQuant
POST /api/indicators?explain=true
{
  "data": {
    "cvd_hist":  [-12, -8, -3, 5, 11, ...],
    "bear_div":  [false, false, true, ...]
  },
  "knowledge": {
    "summary_focus": [
      "Is cvd_hist positive or negative — buyer vs seller dominance",
      "Any recent bull/bear divergence — strength (Normal/Good/Strong)",
      "Whether three timeframes show divergence resonance"
    ],
    "signals": {
      "bear_divergence": "Price makes a higher high but cvd_hist makes a lower high — bull volume waning",
      "strong_signal":   "Strength = 3 — reversal probability rises significantly"
    },
    "caveats": "Divergence needs fractal confirmation (lag); fewer signals in single-direction trends"
  }
}
For your LLM:
  • 📚 Knowledge curated by humans, not invented by the LLM — written by indicator authors, reviewed in the admin console
  • 🎯 `summary_focus` is literally the «look at these things» checklist; the LLM outputs structured analysis by walking 1/2/3
  • 🛡️ `caveats` encode pitfalls the author hit — e.g. fractal-confirmation lag, fewer signals in single-direction trends — preventing over-confident reads

What you actually get

Six out-of-the-box primitives covering AI trading end-to-end — from data ingest to long-term experience. Each is independently callable and freely composable. Pull only the layer you need.
Market data
11 venues, 500+ pairs, 6 timeframes, millisecond WebSocket push with REST fallback. Structured market context, not lonely candles.
In-house indicators
53 TradingView-aligned professional indicators across trend, momentum, volatility, volume, and pattern. Every indicator self-describes, so the AI discovers new capabilities on the fly.
Indicator knowledge base
More than formulas — every indicator carries its applicable regime, combinations, and common traps, so the AI knows when to lean on it and when to ignore it.
Analysis Playbooks
Mature trading logic abstracted into reusable reasoning patterns. Long/short pulses, volatility compressions, volume-price divergences — ready to call.
Multi-platform Skills
Native integrations with Claude Code, Codex, OpenClaw, and Hermes. Wherever your AI agent lives, the capabilities go with it.
Long-term memory
Persistently stores strategy iterations, post-mortems, and personal preferences. The AI grows into you — like raising a double.

Pricing

Four tiers. The production version meters by call volume. The numbers below are a preview of post-beta pricing.
Free
$0
/ forever
5 req/min
All public indicators
Basic Playbooks
Community support
Pro
$29
USD / month
120 req/min
All Playbooks
Long-term memory
Multi-platform sync
Max
$99
USD / month
600 req/min
Private Playbooks
Strategy replay
Priority support
Enterprise
Custom
Contact us
Unlimited calls
Private deployment
SLA-backed
Custom development
Free during beta
All tiers are unlocked, all rate limits lifted — we just want real-world feedback. 💎 Join the «Early Contributor Program» now — one adopted suggestion earns you permanent free access.

FAQ

Plumbing is the first half of the game. Intelligence is the second.
We finish the first half so we can hand the second back to you.
© MobiusQuant