How do I add quant tools to Claude Desktop or Cursor?

Grab a free Kwant API key ($3 in credits, no card) at https://kwant.sh, then paste one MCP server config block into Claude Desktop or Cursor pointing at https://kwant.sh/mcp with an Authorization: Bearer header. Your agent can then backtest strategies, pull indicators, screen and score stocks, and build portfolios over US & TSX equities in plain English.

Can an AI agent backtest a trading strategy?

Yes. Kwant is an MCP server that lets any LLM agent run backtests server-side. Ask a question like 'backtest a 60/40 SPY/AGG portfolio vs SPY 2020–2024, show Sharpe and max drawdown' and the agent calls Kwant's run_portfolio_backtest and compute_portfolio_stats tools and returns structured JSON. Each tools/call costs $0.005 (data/signal) or $0.010 (intelligence).

How much does Kwant cost?

Every new key opens with $3 of free trial credit, no card required — about 600 data/signal queries or 300 intelligence queries. After that, pay per query: $0.005 for data/signal tools, $0.010 for intelligence tools, from a prepaid balance or per call in USDC via x402. Connecting (initialize, tools/list) is always free.

Do I need Python or my own market-data API?

No. Kwant is a hosted MCP endpoint — no install, no Python, no scraping wrappers or data-provider keys. Point any MCP client at one URL and it returns clean, typed quotes, indicators, signals, screens, factor scores, and backtests over US & TSX equities.

Quant Intelligence · MCP over Streamable HTTP · on the edge

Quant intelligence,
as an endpoint.

Backtest a strategy, pull signals, screen and score stocks — ask in plain English and your agent runs it. One URL, no data plumbing, no Python. US & TSX equities.

Start free — $3 in credits See the intelligence

$3 free · no card · works with Claude, Cursor, or any MCP client

endpoint https://kwant.sh/mcp

See it work

A real question, a real result — no signup, nothing to configure.

agent · connected to kwant

Get a free key — email in, key + $3 out

Paste the prompt into Claude or Cursor

Ask market questions — quotes to backtests

Start free — $3 in credits

Drop your email, we'll send an API key with $3 of credits — about 600 data queries or 300 intelligence queries. No card. Fund more later with POST /topup or pay per query with x402.

Check your inbox — we sent your API key + setup instructions to .

⚠ Didn't land in a minute? Check spam, or the key never left our server — try again.

⚠ Shown once — store it now. We keep only a hash and can't show it again.

1 · Add to Claude Desktop or Cursor — paste into your MCP config (key already filled in):

2 · Or drop this prompt into any agent — Claude, Cursor, your own code:

Then paste this into Claude, Cursor, or any agent — swap in the key from your email and it wires up the server and starts querying.

Use the Kwant quant-intelligence MCP server at https://kwant.sh/mcp.
Auth: send the header  Authorization: Bearer YOUR_KWANT_KEY  on every request.

It exposes 16 tools over MCP. `initialize` and `tools/list` are free; each
`tools/call` costs $0.005 (data/signal) or $0.010 (intelligence), billed to my
prepaid balance (I have $3 of free trial credit) or paid via x402.

First call tools/list to see what's available, then use the tools to answer my
market questions — quotes, indicators, signals, screens, factor scores,
backtests, and portfolios over US & TSX stocks. Tickers: US bare (AAPL), TSX
as RY.TO. When you answer, name the tool(s) and tickers you used.

Start by listing the tools and fetching a quote for AAPL.

One key, any number of agents — paste it into Claude, Cursor, or your own code and they all draw on the same balance. See llms.txt for the machine-readable spec.

Three layers

Market data is a commodity. Kwant turns it into judgment — data becomes signal, signal becomes intelligence.

01 · DATA

Clean market data

Point-in-time quotes, OHLCV history, and fundamentals for US & TSX names — typed, cached at the edge.

get_quote
get_price_history
get_fundamentals

02 · SIGNAL

Indicators & signals

Technical indicators, event detection, and risk/return statistics — the readable signals under the noise.

compute_indicator · detect_signals
compute_stats · compare_tickers
compute_correlation_matrix

03 · INTELLIGENCE

Factor scores & strategy

Screen and score whole universes, construct weighted portfolios, and backtest strategies. The quant desk's judgment.

screen · screen_with_scores
compute_universe_scores · construct_portfolio
backtest · run_portfolio_backtest

16 tools · one endpoint

Discover them over tools/list. Priced per query at call time. Show full list ▾

Tool	Does	Layer	Query
`get_quote`	Latest point-in-time quote	data	$0.005
`get_price_history`	OHLCV bars over an interval/range	data	$0.005
`get_fundamentals`	Profile + key ratios (mkt cap, PE, sector…)	data	$0.005
`compute_indicator`	RSI, MACD, SMA, EMA, BBANDS, ATR, ADX, STOCH	signal	$0.005
`detect_signals`	Golden/death cross, RSI bands, MACD cross, breakout	signal	$0.005
`compute_stats`	Volatility, sharpe, max drawdown, returns, beta	signal	$0.005
`compare_tickers`	Rank N tickers by a chosen metric	signal	$0.005
`compute_correlation_matrix`	Pairwise correlation across a basket	signal	$0.010
`screen`	Filter a universe by fundamental/technical rules	intel	$0.010
`screen_with_scores`	Screen + rank with composite factor scores	intel	$0.010
`build_monthly_universe`	Construct a point-in-time monthly universe	intel	$0.010
`compute_universe_scores`	Factor scores across a whole universe	intel	$0.010
`construct_portfolio`	Build a weighted portfolio from a selection	intel	$0.010
`compute_portfolio_stats`	Portfolio-level risk/return statistics	intel	$0.010
`backtest`	Backtest a signal/strategy on one ticker	intel	$0.010
`run_portfolio_backtest`	Monthly-rebalanced multi-asset backtest	intel	$0.010

Built for humans & agents

For the quant dev

Skip the data plumbing — no scraping wrappers, pandas, or cron jobs to babysit. One endpoint returns clean, typed intelligence, from a single quote to a full monthly-rebalanced portfolio backtest. Start free with $3.

For the agent

A first-class MCP tool surface. Attach the key its human provisioned, discover tools over tools/list, query them with JSON args, and draw on the prepaid balance — or pay per query in USDC via x402. The handshake carries no charge.

Pricing & payment rails

Start free with $3 in credits. After that, per-query and metered at the edge: $0.005 data/signal, $0.010 intelligence.

Prepaid — start free, top up by card

Every key opens with $3 of trial credit — no card. When it runs out, top up a balance with a Stripe Checkout link (POST /topup) and each query debits it. Watch it at GET /balance. Best for a human developer running an agent.

$3 free to start · then $0.005 / $0.010 per query from balance

x402 — pay in USDC

Per-query crypto payment settled on-chain (Base mainnet, real USDC via the CDP facilitator) — no API key, no signup. The 402 carries the requirements in a base64 PAYMENT-REQUIRED header (x402 transport v2); retry with PAYMENT-SIGNATURE and get the tx hash back in PAYMENT-RESPONSE. Built for autonomous agents with a wallet.

$0.005 data/signal · $0.010 intelligence — per query

An unpaid priced query returns a unified HTTP 402 advertising every enabled rail (Accept-Payment: x402, prepaid). Capability negotiation is never charged.

For agents

If you are an autonomous agent or a crawler, everything you need is machine-readable. Your human operator provisions the key (start free); you consume the endpoint with it:

`/llms.txt`	Plain-text capability + endpoint + pricing summary for LLMs
`/.well-known/ai-plugin.json`	Plugin-style manifest pointing at the MCP endpoint
`/mcp`	The MCP endpoint. Pay per call with x402 or attach a Kwant API key; `initialize` + `tools/list` carry no charge
`/sitemap.xml` · `/robots.txt`	Crawl map. All reputable AI crawlers allowed

With the key your operator provisioned, discover tools dynamically, price each query against the table above, and draw on the balance or settle per call with x402 — no human in the loop at query time.

FAQ

How do I add quant tools to Claude or Cursor?

Grab a free key above ($3 credit, no card), paste one MCP config block pointing at https://kwant.sh/mcp, and ask market questions in plain English.

Can an AI agent backtest a strategy?

Yes — see the demo. Ask, and the agent calls run_portfolio_backtest + compute_portfolio_stats and reads back Sharpe, CAGR, drawdown.

What does it cost?

$3 free to start, no card. Then $0.005 data/signal, $0.010 intelligence per query. Connecting is always free.

Do I need Python or a data API?

No. Hosted MCP endpoint — no install, no Python, no data-provider keys. One URL returns clean typed JSON.

Quant intelligence,as an endpoint.