Discover the Best AI Tools

Local workbench combining quantitative research, document workflows and futures-contest trading tools. Models and market data require configured services; current futures integration is simulated competition trading, and shared deployments do not isolate users’ data.

CLI and MCP tools for composing and backtesting Hyperliquid strategies. The hosted interface is research-only; live operation is handed to the web app, while optional local live controls require explicit enabling and arming.

Connects agents to a hosted Malaysian individual-tax research and draft-preparation service without an API key. It does not file returns; 2026 filing fields remain blocked pending official forms, and taxpayer personal data should not be submitted.

Unofficial Toss Securities MCP integration for Korean and US stock quotes, holdings and orders. Requires Toss credentials; orders use preparation and confirmation steps and can execute real trades.

Experimental intraday trader ranks Nifty-50 stocks with Jev and submits broker orders with code-based sizing and loss limits. Requires broker and gateway credentials; there is no paper-trading mode, Zerodha is not live-verified and real funds can be lost.

Paid agent-review service evaluates proposed trades against supplied portfolio and risk context, with signed decision records. Its verdict is advisory and never blocks a trading bot; payment access and independent validation remain necessary.

MCP market-analysis tools expose order-flow measurements, market-stress indicators and rule-derived trading context. Requires configured feeds and service access; risk scores and action suggestions are analytical outputs, not validated predictions or execution safeguards.

AI Stock combines market data, news, AI research perspectives and strategy screening with paper trading. Research and simulation cover different market sets, and screening depends on configured data providers. Hosted access uses a platform allowance or your own model API. Simulations never place live broker or exchange orders, and AI expert roles are not real people.

Stillos Notary MCP connects agents to hosted claim resolution, signed receipts, sanctions-name screening and corporate distress ranking. Limited free calls and paid Base USDC tools coexist; configured payment credentials enable automatic payment. Distress scores prioritize review, not reliable individual bankruptcy alarms, and signatures do not establish legal certification.

Self-hosted platform for AI-assisted market research, Python strategy backtesting and paper or live trading. Live execution requires provider credentials and can place real orders; backtests do not guarantee returns and frontend licensing differs from the backend.

Locally calculates UK bridging-finance costs, development viability, rental stress tests and stamp duty. Tax tables are dated April 2026 and need maintenance; the hosted service is not yet public.

Exposes financial-data research tools covering filings, ETFs, macro data and DeFi through a hosted service. Free public tools need no account; paid features use OAuth, and the service does not execute investments.

Calculates Canadian adjusted cost bases, capital gains and superficial losses from transaction data. Targets taxable accounts; exchange rates are supplied by users and complex corporate actions remain unsupported.

Provides company financials, quotes, screening and comparisons for equity research through MCP. Requires an API key and tenant ID; deeper research features depend on the subscription tier.

Searches Japanese company data locally through MCP, with optional semantic retrieval. The free dataset covers 192 companies; broader coverage is paid, and the runtime and data have separate licensing from the bridge.

Portfolio tool computes drawdowns and rule-based rebalance suggestions, with optional constrained AI narration and read-only MCP. Limited to USD long-only stocks and ETFs; users maintain the transaction log and data-dependent estimates are not financial advice.

MCP supplies curated US options candidates, historical features and paper-trade outcome data for agent research. Advanced tools require a subscription; it provides evidence rather than picks, and historical opportunity surfaces are not realized investment returns.

MCP wrapper offers compound-interest, loan, retirement, tax and Islamic-finance calculations through calcnook. Results depend on inputs and supported jurisdictions; currency conversion uses supplied rates and financial or compliance conclusions need verification.

MCP server retrieves crypto, stock, index, forex and sentiment data from public APIs without user keys. Availability, rate limits and source delays still apply; it does not guarantee real-time accuracy or profitable trades.

Hosted MCP connector for debt-collection case status, documents, partner messages and submissions. Requires a Debitura account and API key; submitting cases and messages has real legal or financial effects.

Chart Library MCP exposes completed-session market states, historical analogues and published research with sample receipts, limitations and source hashes. Read-only access is available without a key, subject to service limits. Latest means the latest published session, not a live quote; thin evidence and historical ranges are not guaranteed forecasts.

IBANforge validates IBAN structure and looks up bank, BIC and SEPA registry information via API, SDKs and MCP. Coverage and licensing vary; some directory rows are historical or unavailable when self-hosted. Paid calls use keys/credits or x402. It does not verify account existence, ownership or the payee’s name.

A hosted MCP connector researches global stocks, ETFs and authenticated user portfolios; public data can be queried anonymously, portfolio tools require sign-in and draft tools require Pro, while the repository is the connector layer rather than the full data platform.

A local MCP server turns confined CSV data and reviewed strategy drafts into bounded Backtrader runs and reports; it requires the pinned CloudQuant runtime and supports offline backtesting only, with no live orders or arbitrary Python execution.