This read-only MCP server lets you explore the TradingAgentStack directory of AI trading tools, brokers, APIs, frameworks, backtesters, and security tools.
Search tools by keyword, category, and result limit.
Get full sourced details for a specific tool by slug or name.
Compare two tools or brokers side by side on safety controls, paper trading, pricing, and rate limits.
List directory categories with tool counts.
All capabilities are read-only; it does not place trades or access external systems.
Indexes LangGraph as an agent framework entry in the directory, with its official repository link and a full profile page for building trading agents.
Provides sourced profiles of the Massive (formerly Polygon) MCP server and API, covering market data for stocks, options, forex, crypto, futures, fundamentals and news.
Provides a sourced profile of the Robinhood Trading MCP server, covering equities (long), options and crypto trading from a dedicated MCP account, including live-trading status and links to Robinhood's official onboarding docs.
Provides a sourced profile of the TradingView MCP server, which exposes stocks, ETFs, forex, crypto, futures, indices, bonds and macro data to AI agents for watchlists and alerts (no order placement).
This server has a remotely hosted endpoint. Open its connector listing for connection details, authentication requirements, and setup instructions:
A sourced list of AI trading agents, broker MCP servers, market data APIs, Telegram bots, agent wallets and exchanges. Every fact and score is copied from the TradingAgentStack open dataset.
160 tools. Dataset generated 2026-11-02. Newest listing check: 2026-11-02.
Not financial advice. Verify details with each vendor before connecting real money.
Data: data/tools.json · data/tools.csv
License: CC BY 4.0. Credit TradingAgentStack (tradingagentstack.com).
Scores: How the safety score works
Related MCP server: EdgeDepth Research MCP Server
The safety score is a 0 to 100 measure of what a tool can do with your money. It is built from seven sourced inputs: whether it can place live orders unattended, whether order approval exists, whether paper, demo or testnet is available, whether it can withdraw or transfer funds, whether keys can be scoped, whether the code is open source, and whether the facts come from the vendor. When an input is missing, the listing is marked partial and the percentage is the points earned out of the points that could be checked. Affiliate links never affect the score.
Crypto agent tools (agent wallets, DeFi and Telegram bots, exchanges, and similar) may also show an Agent Risk Score. That column is filled only where TradingAgentStack has published a v2 score. The percentage is the published points divided by the points that could be checked, rounded to the nearest whole number.
See Trade with a bot.
AI trading agents on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
tokens on BNB Smart Chain, Ethereum, Base and Solana, market swaps and limit orders, prediction markets | live on BNB Smart Chain, Ethereum, Base and Solana; no testnet mentioned | official | 65% | 65% | |
swaps and bridges on Ethereum, Solana, BNB Chain, Base, Arbitrum, Tron and more, gasless transfers, x402 payments | live; no testnet mentioned | official | 55% | 55% | |
swaps, bridges and transfers from EVM, Solana and TON wallets, automated strategies on linked exchange accounts | live; no testnet or paper mode mentioned | official | 56% (partial) | 56% (partial) | |
crypto spot and whole-share US equities on Robinhood (long-only, limit orders), DEX swaps through 1inch on 10 EVM chains, Kraken (orders through Mangrove's keyless OAuth connection) | paper schedules run on MangroveAI with no funds at risk; live execution on Robinhood, 1inch DEX swaps and Kraken OAuth | official | 58% | 58% | |
tokens on nearly 20 chains incl. X Layer, Ethereum and Solana, trading and transfers | live (launched March 18, 2026); no testnet mentioned | official | 25% | 25% | |
perpetual futures on Hyperliquid, perpetual futures on Decibel | live only on mainnet; no testnet or practice account in the documented setup | official | 35% | 35% |
AI trading bot platforms on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
crypto spot, crypto futures | live via exchange API keys; free trial; backtesting | official | 58% | ||
crypto spot, crypto futures | live via exchange API keys; free demo with 20 bots | official | 50% | ||
forex, CFDs, crypto, stocks (via partner brokers) | live via partner brokers; backtesting and simulation | official | 33% (partial) | ||
crypto, US stocks, ETFs | live via exchange/broker connection; backtesting; templates | official | 50% | ||
US stocks, ETFs | live in a Composer brokerage account; unlimited backtests | official | 29% (partial) | ||
crypto | live via exchange API or OAuth; paper trading bot included | official | 50% | ||
crypto spot (live on KuCoin, Bybit, Bybit EU, Bitget and HTX), crypto futures and long/short strategies (paper only) | every account starts in paper mode; live spot trading through your own exchange account is optional | official | 53% (partial) | ||
US stocks and options (option spreads up to 4 legs), crypto | free paper trading (Nyria Paper, Alpaca Paper, Tradier Sandbox); live on Charles Schwab, Tradier, tastytrade and Alpaca | official | 51% (partial) | ||
US options, US stocks | live and paper via connected brokers | official | 53% (partial) | ||
crypto spot, crypto futures | live (Pionex is the exchange; bots are built in) | official | 24% (partial) | ||
US stocks, ETFs, crypto | paper trading on Core; live via connected brokers or Quantbase account | official | 50% | ||
US stocks, ETFs, crypto, forex | virtual and brokerage AI robots | official | 33% (partial) | ||
US stocks, ETFs | live via connected brokers; real-time paper trading | official | 33% (partial) | ||
US stocks, futures, options, crypto | paper auto-submit free; live auto-submit on paid plans | official | 72% (partial) | ||
US stocks, ETFs, futures, options, forex, crypto | live via SignalStack to 30+ brokers and exchanges (5 free automated orders/mo); backtesting and forward testing | official | 17% (partial) | ||
crypto spot, crypto futures | live via exchange API keys; demo account | official | 50% |
Open-source trading bots on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
crypto spot, crypto futures | dry-run (paper) and live, self-hosted | official | 60% | ||
crypto spot, crypto perpetuals, DEX, Kalshi | live (self-hosted); paper connectors | official | 60% | ||
crypto, stocks, commodities | backtesting and live (self-hosted) | official | 53% (partial) | ||
stocks, ETFs, options, futures, forex, crypto, prediction markets | backtest, paper and live (self-hosted); the strategy class can stay the same, but startup code must call Strategy.backtest() for history or run_live()/Trader.run_all() for a broker. Alpaca paper is PAPER true or ALPACA_IS_PAPER, and the CLI has lumibot run --paper. Alpaca, Tradier, Interactive Brokers and Tradovate document paper or demo accounts; CCXT can use sandbox true where the exchange has a testnet. Setting IS_BACKTESTING=false does not by itself turn a backtest-only file into broker trading. | official | 53% (partial) | ||
crypto | live and paper (virtual money) | official | 60% | ||
crypto | backtest, paper and live (self-hosted) | official | 60% | ||
stocks (US, HK, China A-shares, UK, Canada, India, Korea, Brazil and more), crypto, futures, forex | backtest, paper and live (self-hosted). v0.1.17 (Oct 21, 2026) adds daily paper trading with saved accounts that fill against completed daily bars and never reach a broker. Most of its 19 broker connectors also place orders on the broker's own paper account; bounded live orders are wired for Alpaca, Binance, OKX, Futu, Tiger, eToro, MetaTrader 5 and Robinhood (Robinhood is live-only). Paper vs live is a per-broker runtime check (account id, host, demo flag), not a setting the agent can flip, and brokers without such a signal are capped at paper and read-only. | official | 62% (partial) |
On-chain & DeFi trading agents on TradingAgentStack
Guide: Best Telegram trading bots.
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
EVM chains, Solana | live (on-chain swaps) | official | 41% (partial) | 41% (partial) | |
EVM DEX liquidity | live (on-chain swaps) | official | 71% (partial) | 71% (partial) | |
Solana, Ethereum, Base, other EVM chains | live only (real funds) | official | 11% (partial) | 11% (partial) | |
Solana | live only | official | 11% (partial) | 11% (partial) | |
Solana tokens (any token with a liquidity pool, including new pump.fun coins), routed through Jupiter, cross-chain swaps (Bitcoin, Ethereum, Base, BNB Chain, Arbitrum and more) through LI.FI, token launches on 8 chains | live only: real tokens on Solana and other mainnets; no paper or demo mode on the site | official | 80% | 80% | |
Solana DEX liquidity | live (on-chain swaps) | official | 41% (partial) | 41% (partial) | |
Solana, Ethereum, BSC, Base, Arbitrum, TON, Tron | live only | official | 10% | 10% | |
Solana | live and devnet | official | 35% | 35% | |
Solana | live only | official | 10% | 10% |
AI signals & research copilots on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
US stocks, ETFs, European stocks | research only (no execution); broker sync on paid plans | official | 90% | ||
US stocks, ETFs, crypto | research only | official | 90% | ||
crypto | research only | official | 90% | ||
US stocks, ETFs, global stocks and ETFs (30+ exchanges in 29 countries, Pro and Premium), crypto, forex | research only | official | 90% | ||
US stocks, options | alerts only | official | 90% | ||
crypto, stocks | research only | official | 90% | ||
crypto (on-chain) | research only | official | 90% | ||
US stocks | research only | official | 90% | ||
crypto | research only | official | 90% |
Copy trading & strategy marketplaces on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
crypto futures, spot | live | official | 29% (partial) | ||
crypto derivatives, spot | live | official | 51% (partial) | ||
stocks, options, futures, forex, crypto | live AutoTrade at supported brokers; strategies tracked publicly | official | 41% (partial) | ||
CFDs, futures, crypto CFDs | virtual capital; track record can attract real allocation | official | 90% | ||
stocks, ETFs, crypto | live; free $100k virtual portfolio for demo | official | 51% (partial) | ||
forex, CFDs (MetaTrader 4/5) | live only: copyable signals must come from real-money accounts (demo, contest and cent account signals are not allowed) | official | 43% | ||
forex, CFDs, crypto CFDs | live and demo | official | 44% (partial) |
Agent tokens with a product on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
prediction markets (Polystrat on Polymarket, Omenstrat), DeFi portfolio agents (Basius on Base, Optimus) | live | official | 33% (partial) | 33% (partial) | |
USDC agent-to-agent job payments held in escrow, agent wallets (EVM and Solana), virtual payment cards | live on Base mainnet (ACP v2.0, April 2026); Base Sepolia and BSC testnet also supported | official | 95% | 95% |
See Build an AI agent.
Broker MCP servers on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
US stocks, ETFs, options, crypto, fixed income quotes | live and paper (ALPACA_PAPER_TRADE defaults to true/paper; paper with simulated $100K) | official | 68% | ||
Axi MT5 account instruments (live prices, candles, positions, orders) | live and demo MT5 accounts; MT4 and Axi trading platform accounts can't connect | official | 75% | ||
forex, crypto spot and crypto perpetual futures, CFDs on indices, commodities, energies and metals, NDFs | live: the full-access endpoint places real orders on your B2Trader account | official | 40% (partial) | ||
crypto spot and derivatives on 100+ exchanges via the unified ccxt API, prediction markets (e.g. Polymarket) | live and sandbox/testnet accounts (per exchange); out of the box only public market data; trading must be enabled per account in the config file, and | official | 85% | ||
forex, CFDs (instruments depend on your cTrader broker) | live: with trading enabled, the agent places, modifies and closes real orders on the connected cTrader account | official | 64% (partial) | ||
stocks, ETFs, crypto, commodities | live account; eToro suggests testing workflows on a demo account first | official | 90% | ||
global multi-asset (account/portfolio, market data) | live account | official | 71% (partial) | ||
quotes/depth, fundamentals, options, warrants | live trading tools (place/modify/cancel); paper trading accounts work too, except on Longbridge US accounts, which have no paper account | official | 68% | ||
quotes, portfolio, orders; HK, US, CN, SG, JP, MY markets (OpenAPI) | Simulated (paper) trading is authorized separately from live trading; paper trading becomes available with market data authorization, so an agent can use paper trading without live trade access. moomoo says its docs may currently be wrong on this (moomoo OpenAPI team, email to TradingAgentStack, 2026-10-26) | official | 58% | ||
futures and event contracts in your NinjaTrader or Tradovate account (contracts, quotes, orders, risk) | Demo (simulated) and Live endpoints | official | 58% | ||
Public.com brokerage account | live (real money); no paper or sandbox account: every order is a real order in your Public account (Public, email to TradingAgentStack, 2026-10-26) | official | 28% (partial) | ||
equities (long), options, crypto | live, from a dedicated MCP account | official | 71% (partial) | ||
tastytrade account (orders, quotes) | live by default (connects to production unless TASTYTRADE_ENV=sandbox); sandbox available (no market data, resets every 24 hours) | official | 100% | ||
TradeStation execution, market data, account APIs | live | official | 86% (partial) | ||
US stocks, options (single-leg and multi-leg) | live and paper (PAPER_TRADING flag with your paper token) | official | 50% | ||
stocks, ETFs, options, futures, crypto, event contracts, mutual funds | live: creates order instructions (stocks, options, futures, crypto, event contracts, algo) that you confirm in the Webull app/desktop; plus account data, order history, market data | official | 88% (partial) |
Market data MCP servers on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
stocks (20+ yrs history), quotes | data only | official | 100% | ||
Bitcoin, Ethereum, XRP, TRON, stablecoins, ERC-20 tokens, alts | data only | official | 90% | ||
Databento datasets | data only | community | 100% | ||
DEX tokens and pools, on-chain crypto across many chains | data only | official | 100% | ||
China A-share stocks and concept sectors (news-flow heat: mentions, institutions and influencers mentioning them, first mentions), optional Hong Kong and US names by stock name | data only | official | 90% | ||
stocks, options, forex, crypto, futures, fundamentals, news. Formerly: Polygon MCP (mcp_polygon); repo renamed from polygon-io/mcp_polygon to massive-com/mcp_massive | data only | official | 100% | ||
crypto, FX, equities, metals, rates, commodities, funding rates, NAVs | data only | official | 90% | ||
about 2,900 US-listed stocks and 1,200 US ETFs (end-of-day returns and risk metrics, up to 5 years), portfolio backtests with quarterly rebalancing | data only | official | 90% | ||
US stock and ETF options flow (captured 0-2 DTE activity for 100 tracked tickers), large options prints ($1M+) out to 45 DTE, daily open interest by strike | data only | official | 90% | ||
US stocks, forex, crypto, commodities, DEX and on-chain data, SEC fundamentals | data only | official | 100% | ||
stocks, ETFs, forex, crypto, futures, indices, bonds, macro | no trading (watchlists/alerts only) | official | 90% |
Agent frameworks on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
n/a | n/a | official | 100% | ||
n/a | n/a | official | 100% | ||
n/a | n/a | official | 100% | ||
n/a | n/a | official | 100% | ||
stocks (any Yahoo Finance ticker: US, HK, Tokyo, London, India, China A-shares and more), crypto (e.g. BTC-USD) | research and backtesting only; approved trades go to a simulated exchange (no broker execution built in) | official | 100% |
Risk, safety & paper trading on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
crypto, stocks | live market data, paper (simulated) trading: the repository's results are paper_trading files | academic | 100% | ||
US stocks, ETFs, options, crypto | paper (simulated) with the same API as live | official | 90% | ||
MCP servers (any) | n/a | third-party | 90% | ||
LLM agents (any) | n/a | official | 100% | ||
stocks, options, futures, forex | paper with $1M simulated buying power | official | 90% | ||
MCP and agent tool calls | n/a | official | 100% | ||
LLM agents (any) | n/a | official | 100% | ||
orders from any bot or agent | n/a | community | 100% | ||
stocks, forex, crypto, futures, commodities | paper | official | 90% |
AI agent wallets on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
EVM chains, Solana | live and testnet | official | 65% | 65% | |
EVM networks (e.g. Base) | live and testnet (the quickstart runs on Base Sepolia with CDP Faucets) | official | 55% | 55% | |
stablecoins (e.g. USDC), on-chain payments | live and staging (staging is connected to testnets, funded with test tokens of no value) | official | 60% | 60% | |
Ethereum and EVM chains, Solana | live and testnet (the REST quickstart sends a Sepolia transaction) | official | 65% | 65% | |
EVM chains | live; contracts are deployed on many networks (see Safe's supported-networks list) | official | 100% | 100% | |
EVM chains, Solana, Bitcoin, Dogecoin, Tron, Cosmos and other chains | live; official code examples sign on testnets (Dogecoin, Tron, Celestia, Arc, Tempo) | official | 65% | 65% | |
EVM chains (ERC-4337 smart accounts) | live and testnet (the free Sandbox plan is testnet only) | official | 70% | 70% |
AI agent payments on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
card payments (initial version), x402 stablecoin sample | specification, SDK and code samples (v0.2); not a hosted payment service | official | 75% (partial) | 75% (partial) | |
USDC on Base, Polygon and Solana, ETH, POL and SOL sends, token swaps on Base and Polygon, x402 paid APIs | live; the CLI also supports Base Sepolia and Solana Devnet | official | 55% | 55% | |
bitcoin over the Lightning Network, L402 pay-per-request APIs | mainnet, plus regtest and testnet setups (local-key mode is meant for testnet, regtest and small-value mainnet) | official | 56% (partial) | 56% (partial) | |
one-time-use cards from your Link wallet, Shared Payment Tokens for Machine Payments Protocol (MPP) merchants | live; US Link accounts only (link-cli README); link-cli can issue test Shared Payment Tokens for a Stripe sandbox | official | 75% | 75% | |
virtual and single-use cards, Shared Payment Tokens for MPP and UCP checkouts, agentic network tokens (Visa, Mastercard) | live, by application (apply in the Stripe Dashboard; platforms contact Stripe); the agents page doesn't mention a test mode | official | 55% | 55% | |
USDC and other ERC-20 tokens on Base, SPL tokens on Solana, pay-per-request HTTP 402 payments | live on Base and Solana mainnet; testnets on Base Sepolia and Solana Devnet | official | 64% (partial) | 64% (partial) |
Exchanges & brokers for bots on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
crypto perpetuals, crypto spot | live; testnet app and API with a daily faucet | official | 58% | 58% | |
crypto spot, crypto margin, crypto futures | live; Spot Testnet (testnet.binance.vision) for paper testing | official | 49% | 49% | |
crypto spot, crypto futures, copy trading | live; demo trading in app, and the --paper-trading flag routes MCP, CLI and SDK requests to Bitget's Demo environment | official | 85% | 85% | |
crypto spot, crypto derivatives, options | live; Demo Trading (api-demo.bybit.com, live prices) and a separate Testnet | official | 85% | 85% | |
crypto spot, crypto futures (eligible regions), crypto perps (20+ tokens) and index perps (US only) via Coinbase for Agents, eligible US stocks via Coinbase for Agents | live; static Advanced Trade sandbox (mocked Accounts and Orders responses) | official | 75% | 75% | |
crypto perpetuals | live; public testnet with faucet | official | 68% | 68% | |
crypto perpetuals | live | official | 29% (partial) | 29% (partial) | |
crypto perpetuals, RWA perpetuals | live; Sepolia testnet with daily test USDC | official | 43% | 43% | |
crypto spot, crypto perpetuals (eligible regions), prediction markets | live; full-function sandbox with test funds (exchange.sandbox.gemini.com) | official | 60% | 60% | |
crypto perpetuals (pool-based, oracle-priced) | live; testnet contracts on Arbitrum Sepolia and Avalanche Fuji | official | 68% | 68% | |
crypto perpetuals, crypto spot (on-chain) | live; public testnet (api.hyperliquid-testnet.xyz) | official | 50% | 50% | |
crypto spot, crypto futures, xStocks, forex | live; built-in spot and futures paper trading in Kraken CLI (no API key needed) | official | 85% | 85% | |
crypto perpetuals, crypto spot | live; public testnet (testnet.app.lighter.xyz, settles to Sepolia) | official | 68% | 68% | |
crypto spot, crypto perpetuals, futures, options | live; demo trading profile (simulated) supported by the MCP | official | 85% | 85% | |
crypto perpetuals, crypto spot | live; testnet API (the official MCP server defaults to testnet) | official | 62% (partial) | 62% (partial) | |
crypto perpetuals, crypto options, crypto spot | live; testnet with $1M test USDC per account | official | 68% | 68% | |
perpetuals on crypto, US stocks and indices, FX, gold and oil | live on Base; testnet paper trading with test USDC | official | 75% | 75% |
Prediction market APIs & agents on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
Kalshi 15-minute up/down markets: 9 crypto (BTC, ETH, XRP, SOL, DOGE, HYPE, ZEC, NEAR, BNB), 5 metals, oil, natural gas and 3 currency pairs, Kalshi multi-leg combos and parlays, Polymarket 15-minute crypto Up/Down markets (on your own Polymarket account) | live on your own Kalshi account (and optionally Polymarket); starts in paper mode with simulated trading until you arm each market | official | 44% (partial) | ||
prediction markets, perpetual futures (Margin/perps markets) | live; separate demo environment with mock funds (demo.kalshi.co) | official | 50% | ||
prediction markets | demo and live; trading off by default | community | 85% | ||
prediction markets (crypto, on Base) | live only; no sandbox or testnet (docs say rehearse with minimum-size live orders) | official | 100% | ||
play-money prediction markets | play money (mana); no real-money risk | official | 90% | ||
prediction markets | live (trades via the Polymarket API). Status: Archived by Polymarket on GitHub (read-only). The last code change was Nov 5, 2024, before Polymarket's switch to CLOB V2 on Apr 28, 2026, so treat it as reference code, not a ready-to-trade agent. | official | 45% (partial) | ||
prediction markets, perpetual futures via the separate Perps API | live only (no testnet); CLOB V2 since Apr 28 2026 | official | 30% | ||
prediction markets (US) | live | official | 67% (partial) | ||
prediction markets (Kalshi, Polymarket, Limitless) | data only (no execution) | third-party | 100% |
Broker APIs on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
US stocks, ETFs, options, crypto | live and paper | official | 50% | ||
global stocks, options, futures, forex | live and paper | official | 41% (partial) | ||
US stocks, options | live | official | 17% (partial) | ||
US stocks, options | live and paper (sandbox) | official | 41% (partial) |
Market data APIs on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
stocks, ETFs, mutual funds, indices, options, forex, commodities, economic indicators, news & sentiment | data only | official | 90% | ||
crypto derivatives: liquidations, open interest, funding rates, long/short ratios, options, Hyperliquid positions and vaults, ETF flows, stablecoins, exchange balances | data only | official | 90% | ||
US stocks (trade setups) | data only (no order placement) | official | 90% | ||
US equities, US futures, US options, crypto, FX | data only | official | 90% | ||
stocks, options, indices, forex, futures | data only | official | 90% | ||
multi-provider wrapper | data only. Status: Company winding down (announced Aug 25, 2026). Code released under Apache 2.0 and maintained by the non-profit OpenBQ (github.com/openbq-org). | official | 100% | ||
prediction markets (Polymarket): crypto, sports, esports and weather markets | data and research only | community | 90% | ||
US filings/XBRL fundamentals | data only | official | 90% | ||
stocks, options, indices, interest rates | data only | official | 90% | ||
stocks, options chains | data only | community | 100% |
Backtesting on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
any (bring data) | backtest only | official | 100% | ||
multi-venue incl. IBKR and crypto | backtest and live | official | 47% (partial) | ||
equities, forex, futures, options, crypto | backtest, paper, live | official | 60% | ||
any (bring data) | backtest/research only | official | 90% |
VPS & hosting for bots on TradingAgentStack
Tool | Markets | Live or paper | Maintainer | Safety | Agent risk |
Python/Node bots | n/a | official | 90% | ||
forex (MT4/MT5) | n/a | official | 90% | ||
Python/Node bots (Freqtrade, Hummingbot) | n/a | official | 90% | ||
forex (MT4/MT5) | n/a | official | 90% | ||
Python/Node bots | n/a | official | 90% | ||
futures, forex, crypto bots | n/a | official | 90% |
tradingagentstack-mcp (npm package tradingagentstack-mcp): read-only MCP server for this directory. Hosted endpoint: https://tradingagentstack.com/api/mcp. Source: intoApex/tradingagentstack-mcp.
TAS Guard: open-source limits, approvals and a kill switch between an AI agent and a broker MCP server or wallet. Source: intoApex/tas-guard.
TAS Ratings: the same open data, including Agent Risk Scores, incidents and guard profiles, under CC BY 4.0.
A small stdio server in mcp-server/ reads data/tools.json from this repo.
The tool records in this repository come from the public dataset at tradingagentstack.com/data:
data/tools.json is tools.json, with the same records and indentation added so the file is reviewable. It includes every tool, safety inputs, Agent Risk Scores where they exist, the incident list and TAS Check records.
The same JSON is mirrored in intoApex/tas-ratings. .github/workflows/sync-dataset.yml runs scripts/sync_from_site.py every Monday and commits only when the dataset or this README changes.
The dataset is licensed under CC BY 4.0. Credit TradingAgentStack (tradingagentstack.com) and link to the site. Not financial advice.
This list is regenerated from the public dataset, so corrections belong upstream. Open an issue here, or send the official source through tradingagentstack.com/submit. See CONTRIBUTING.md.
Side-by-side comparison of two brokers/tools (safety controls, paper trading, pricing, rate limits).
| Name | Required | Description | Default |
|---|---|---|---|
| a | Yes | First tool slug or name | |
| b | Yes | Second tool slug or name |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint=true and openWorldHint=false, so the safety profile is covered. The description adds which attributes get compared, but says nothing about behavior for unknown slugs, whether the comparison is symmetric, or how the result is presented.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
A single tight sentence with the resource front-loaded ('Side-by-side comparison of two brokers/tools') and the scope in a compact parenthetical. Nothing redundant.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
No output schema exists, but the parenthetical list of comparison dimensions effectively previews what the agent gets back, and both required params are documented. Minor gap: no handling notes for invalid or identical inputs.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100% — both a/b are documented as 'First/Second tool slug or name.' The description implies the two arguments are brokers/tools but adds no format, slug, or ambiguity-resolution detail beyond the schema, so baseline 3 applies.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb+resource: a side-by-side comparison of exactly two brokers/tools, and enumerates the dimensions compared (safety controls, paper trading, pricing, rate limits). It distinguishes itself implicitly from get_tool (single tool) by specifying two inputs, though it never names the sibling.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description says what it compares but gives no guidance on when to choose it over get_tool (e.g. 'use get_tool for one tool, compare_brokers for two'). The 'two' arity hints at usage but no exclusions, prerequisites, or alternatives are stated.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Full sourced record for one tool: assets, live vs paper, order approval/safety controls, pricing, rate limits, source and last-checked date.
| Name | Required | Description | Default |
|---|---|---|---|
| slug | Yes | Tool slug or name, e.g. 'alpaca-mcp-server' |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare the safe read-only and closed-world profile, so the bar is lower. The description adds valuable uncovered context by listing what the record contains (assets, live vs paper, approval/safety controls, pricing, rate limits, source, last-checked date) and signals data provenance/freshness, which goes beyond the annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
A single dense sentence with the core purpose front-loaded and no filler. Every listed field contributes useful information about the return payload.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
With no output schema and a fully documented single parameter, the description does the necessary work of enumerating return fields. It is largely complete, though it does not describe the response shape or behavior when the slug is unknown.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
There is one parameter, and schema coverage is 100% with an example value ('alpaca-mcp-server') in the schema itself. The description adds no additional parameter meaning, so the baseline of 3 is appropriate when the schema fully documents the input.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description gives a clear verb+resource ('Full sourced record for one tool') and enumerates the record contents, so an agent knows this is a single-item detail lookup. It does not explicitly name the sibling it contrasts with (search_tools), but 'for one tool' implies the distinction.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Usage is only implied by the singular scope: fetch details for one identified tool. There is no explicit when-to-use guidance, no statement of prerequisites, and no routing to alternatives such as search_tools or compare_brokers.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List directory categories with tool counts.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The annotations already declare readOnlyHint=true and openWorldHint=false, so the safety profile is covered. The description usefully adds that results include tool counts, but says nothing about ordering, pagination, or empty-state behavior, and there is no output schema to fall back on.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
A single eight-word sentence with no filler, front-loading the action and the resource immediately. Nothing is wasted.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a no-argument listing tool with no output schema, the description conveys the essential return content (categories, each with a tool count). Ordering and pagination details are absent, but they are minor for a small taxonomy listing.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The tool takes zero parameters, so the baseline is 4. The description correctly implies no filtering or input is needed, and there is no schema surface left unaddressed.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb (List) and resource (directory categories), plus the payload shape (with tool counts). It is clearly distinct from siblings like search_tools and get_tool, though it does not explicitly name or contrast them.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Usage is only implied: a zero-parameter enumeration tool named list_categories is self-evidently for browsing the category taxonomy. There is no explicit when-to-use statement, no mention of alternatives such as search_tools, and no exclusions.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Search the TradingAgentStack directory of broker MCP servers, broker/market-data APIs, agent frameworks, backtesters and security tools. Returns short summaries; use get_tool for full details.
| Name | Required | Description | Default |
|---|---|---|---|
| limit | No | ||
| query | No | Keywords, e.g. 'options paper trading' or 'crypto' | |
| category | No | Optional category filter (see list_categories) |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint=true and openWorldHint=false, so safety profile is covered. The description adds that results are short summaries, which is useful. However it doesn't mention pagination behavior beyond the schema's limit, or result ordering, so it adds only modestly beyond annotations.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Two tight sentences, front-loaded with the resource scope, with the routing to get_tool in the second. No wasted words.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Complete enough for a read-only search tool with annotations covering safety and schema covering two of three parameters. The absence of an output schema means the return format is only loosely described ('short summaries'), leaving some inference to the agent.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 67%; query and category are documented in the schema, and the description reinforces semantics by describing the return type but adds no new parameter detail. Baseline 3 is appropriate since the schema does most of the lifting.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb (Search) and a specific resource (the TradingAgentStack directory), and enumerates the content types it covers. An agent can immediately tell it apart from get_tool and list_categories.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicitly states the alternative: 'use get_tool for full details', establishing the summary-vs-detail split. It also directs the category filter to list_categories. Clear when-to-use-this vs siblings.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.1.0compare_brokersget_toollist_categoriessearch_toolsScored across 4 tools
Each tool maps to a clearly distinct action and resource: search summaries, full record retrieval, pairwise comparison, and category listing. The only mild overlap is compare_brokers potentially involving 'tools' as well as brokers, but the description resolves this clearly.
All four names follow a consistent snake_case verb_noun pattern: search_tools, get_tool, compare_brokers, list_categories. The convention is predictable and readable throughout.
Four tools are well-scoped for a read-only directory/lookup server. Search, detail, comparison, and category listing cover the essential browsing workflow without redundancy.
The surface covers the core discovery workflow: finding tools, getting full details, comparing options, and listing categories. Minor gaps exist around direct browse-by-category, filters, or pagination, but agents can work around these via search.
Read-only search of the Agent Connectors MCP directory.
Search MCP servers, MCP clients and AI agents, and retrieve listing details. Free, read-only access.
Search a curated directory of agent tools, MCP servers, APIs, protocols and skills. No auth.
Search MCP servers, agent skills and open-source AI projects, plus AI tools. Read-only.
