QuantElse
中文 Choose with AI Explore strategies
Product Documentation

From the web to your harness,
use one strategy language.

This public documentation is for product users. Start in the marketplace, or use structured APIs, CLI commands, and the local Runner with your AI harness.

01

Product overview

The strategy marketplace centers on each strategy's version, market, timeframe, risk, parameters, and evidence. Data, analysis, scheduling, and execution frameworks are runtime compatibility boundaries: after a user chooses local or hosted, the system matches the required Runtime. Planned entries expose product and parameter space without pretending to run.

Current status

Technical strategies have runnable configs and arbitrage has selected monitoring configs. Options and Polymarket are discoverable but not runnable. Real historical datasets, public hosted self-service, and production live remain unavailable.

02

Using the web

  1. Browse strategies directly, filtering by market, category, and currently available runtime.
  2. Read the thesis, risks, stop conditions, parameters, and evidence status.
  3. For an implemented strategy, choose hosted or local and let the system match its runtime components.
  4. Planned entries have no fake runtime or download action.
03

No-install AI Context

Give the public AI Advisor Context to Codex, Claude, OpenClaw, or another harness. The server-rendered page exposes preference cards, the current strategy catalog, versions, evidence status, and a recommendation endpoint without requiring a CLI or Skill.

  • In Codex, switch to Plan mode and copy the dedicated prompt to request native single-select cards with two or three options and conditional market questions.
  • The page can request cards, but the current client and question tool decide whether to render them; text is the fallback.
  • Recommendations cite a published strategy slug/version.
  • A market with no published strategy returns an empty result, not a crypto substitute.
  • The public layer stores no session, accepts no secrets, and places no order.
04

Hosted runtime

Hosted runtime is for 24/7 operation, remote visibility, and low-frequency shared signals. Paper is the default. Limited live spot requires a server switch, a validated encrypted account, and explicit bot activation.

PaperNo API keyDefault modeValidate signals and flow
Live · limitedTrade-only, no-withdrawal keyAdministrator closed testSpot, 1×, order cap

The website does not yet offer public self-service. Identity, tenant authorization, automatic reconciliation, a position ledger, and a kill switch remain prerequisites.

05

Local Runtime

Local mode requires no signup and runs multiple compatible strategies in one background process. The download center now combines a framework directory, each framework's strategy configurations, and verified release files. The current packaged Runtime and Developer Source belong to Technical Indicators. Future options, Polymarket, or multi-leg frameworks may have separate dependencies and config formats instead of pretending the current Runtime is universal.

terminal
quantelse-local validate --config runner.json
quantelse-local run --config runner.json --offline
quantelse-local run --config runner.json --ui --offline
quantelse-local context --bot my-strategy

Strategy JSON never contains API keys. Local credentials enter the OS keyring through the terminal. AI context includes redacted state, signals, and events. Hyperliquid public candles can drive paper execution; live still requires a validated account, a global switch, and explicit activation.

06

CLI / AI harness

Codex, OpenClaw, and other harnesses should call controlled commands and APIs. They should not read trading secrets or execute arbitrary AI-generated trading code. The professional strategy-management CLI/Skill is the second layer and remains planned.

Recommend a strategy

terminal
quantelse recommend \
  --goal 'conservative BTC grid' \
  --budget 1000 \
  --risk low

Harness contract

  • Natural-language intent becomes a validated strategy specification.
  • Backtest output carries the data source, costs, and warnings.
  • Runner commands emit structured single-line JSON.
  • AI receives strategy and bot references, never secrets.
07

HTTP API

Public Context, recommendations, and backtests do not require a CLI. See the full interactive API Reference ↗.

GET/v1/frameworksRuntime compatibility for developer tools
GET/v1/strategiesTag, market, risk, and status filters
GET/v1/strategies/{slug}Detail, parameter space, and evidence
GET/v1/advisor/contextDiscovery snapshot; recommendations use available entries only
POST/v1/backtestsBacktests for supported technical strategies
08

Backtest data

Published performance starts empty. Technical demo backtests use synthetic data and label it demo-paper-synthetic. Funding and multi-leg, options, and prediction-market strategies require synchronized order-book, options-scenario, and event-resolution engines respectively. Results stay empty until the matching engine is released.

09

API key safety

  • Paper mode needs no exchange credentials.
  • Live keys never belong in URLs, parameters, logs, or AI chat.
  • Disable withdrawal and transfer permissions; prefer subaccounts and IP allowlists.
  • Hosted credentials are encrypted. Local credentials use the operating-system keyring; strategies and AI receive only an opaque account_ref.
Real-money warning

Popularity, backtests, and AI analysis do not guarantee future results. Switching to live must remain a clear, separate, reversible user action.