markets.tally/mcp
11 changes recorded. MCP Harbor re-reads this listing from the official registry every six hours and asks the live server what it exposes, and keeps what changed between one look and the next.
https://tally.markets/llms.txt
− Free: 3 active theses, including drafts, and 10 total created. Closing or abandoning frees an active slot but does not reset the total. Saved history remains readable, reviewable and closable. Sentinel: $4.99/month or $49/year, with no creation cap and supported monitoring across the active journal. Monitoring is limited to supported data sources. Every plan, including Free, includes AI connector access. Connected adds read-only SnapTrade broker snapshots for $89/month. Tally checks hourly; upst+ Free: 3 active theses, including drafts, and 10 total created. Closing or abandoning frees an active slot but does not reset the total. Saved history remains readable, reviewable and closable. Sentinel: $9/month or $49/year, with no creation cap and supported monitoring across the active journal. Monitoring is limited to supported data sources. Every plan, including Free, includes AI connector access. Connected adds read-only SnapTrade broker snapshots for $89/month. Tally checks hourly; upstrea
https://tally.markets/llms.txt
− > Tally (tally.markets) is an AI agent for prop firm challenges. The− > user starts a free trial at Propr, pastes one API key into Tally,− > chooses the markets, and authorizes the agent by name. Tally then− > trades the challenge on fixed rules, with one scan a day after the− > daily close and a tick every hour, a reduce-only protective stop on− > every position resting at Propr, and a halt before Propr's limits. The− > user sees every order on the status screen and in a daily report, and− > stops the agent with one click. Tally never buys a challenge, moves− > funds or requests a payout. Propr accounts are simulated.− Building an AI agent to pass prop firm challenges, hosted.− Category: AI agent for prop firm challenges · AI trading agent for− Propr · automated prop firm challenge trading · hosted trading agent− for Hyperliquid perpetuals.− Where it runs: Propr (propr.xyz), a prop firm for crypto perpetuals on− Hyperliquid. The free trial is a $5,000 simulated account with no time− limit, a 10% profit target, a daily equity limit 3% below the day's opening balance, including open-position− gains and losses, and a 6% static drawdown limit that counts open positions.− Touching either limit closes the account for good. A paid challenge− with a static drawdown (Classic, Turbo or Pro, any size) runs the same− way after a consent that names the fee at risk; the engine reads the− product's limits and keeps its own inside them (a daily halt at two− thirds of the daily loss limit, a full stop at three quarters of the− drawdown, one fifteenth of the drawdown risked per position); every− chosen market runs on a paid challenge too. The 2-Step products use a− trailing drawdown, not supported yet. One account per Tally account for now.− Markets, as presets: crypto majors (BTC, ETH, SOL), all crypto− (plus XRP, DOGE, AVAX, LINK), metals and energy (gold, silver, copper,− platinum, WTI and Brent crude, natural gas), US stocks (Apple, Tesla,− Nvidia, Microsoft, Amazon, Alphabet, Meta, AMD), currencies (euro, yen,− pound), or everything. The non-crypto markets are perpetuals on− Hyperliquid's XYZ builder dex, which trade every hour of every day. Each− chosen market is checked against the user's Propr account at setup and− any Propr does not offer is skipped by name.− The rules, the same for every account: one scan a day after the daily− close; a close above the prior 20 days' highs opens a long and below− the prior 20 days' lows opens a short; a close back through the prior− 10 days' range closes it; each position risks 0.4% of the starting− balance with a stop two ATR from the fill resting at Propr; at most− three positions open, gross exposure never above twice the equity; the− agent stops opening positions for the day at a 2% loss of equity from+ > Run your trading like a business. AI trading and risk management tools for planning, risk rules, evidence, and decision reviews in Claude and ChatGPT.+ Write a falsifiable thesis, risk budget, catalyst window and exit criteria. Review new evidence against the original commitments. Keep process and outcome separate. MCP journal actions never execute brokerage trades.+ Optional rules-based strategy execution on supported simulated Propr challenge accounts is authorized and controlled separately in Tally. Connected adds read-only broker data to the journal; it is not an execution connector.+ Free: 3 active theses, including drafts, and 10 total created. Closing or abandoning frees an active slot but does not reset the total. Saved history remains readable, reviewable and closable. Sentinel: $4.99/month or $49/year, with no creation cap and supported monitoring across the active journal. Monitoring is limited to supported data sources. Every plan, including Free, includes AI connector access. Connected adds read-only SnapTrade broker snapshots for $89/month. Tally checks hourly; upst+ - [Connect your AI](https://tally.markets/connect)+ - [Journal](https://tally.markets/journal)+ - [Pricing](https://tally.markets/#pricing)+ - [FAQ](https://tally.markets/faq)+ - [Support](https://tally.markets/support)+ - [Privacy](https://tally.markets/privacy)+ - [Terms](https://tally.markets/terms)+ - [MCP skill](https://tally.markets/tally-skill.md)+ MCP endpoint: https://tally.markets/api/mcp (OAuth). Call get_started for the current protocol. Tool access depends on client, plan and permissions. Existing optional Propr accounts remain in Settings; the connector only reports their state and cannot execute, resume or modify trading.
Showing the first 40 of each.
https://tally.markets/llms.txt
− > Tally (tally.markets) is anomaly detection for AI trading agents. The− > trader connects the brokerage read-only (through SnapTrade; Robinhood,− > Schwab, Webull, Alpaca, Fidelity, IBKR and 20+ more); Tally records− > every order the agent places, builds the account's own baseline — which− > symbols, what size, how often, at what hours, how often it cancels —− > scans the account against it every hour, and emails the owner the hour− > the agent's behavior deviates from it. Most days it sends nothing.− > Tally never executes trades: all broker access is read-only, by− > architecture and by policy.− Every trading connector makes your AI more capable. Tally makes it− accountable.− Category: anomaly detection for AI trading agents · AI trading agent− monitoring · AI agent risk monitoring for Robinhood Agents · read-only− brokerage monitoring.− Why an AI trading agent goes wrong: it acts on input it cannot verify.− Prompt injection (a page it reads carries instructions for it), a− poisoned data feed, a bad model or tool update, a runaway loop. None of− it is visible in the chat, where the agent explains every trade− fluently, compromised or not. It is visible in the account, as an order− the agent would never have placed the day before. That is what Tally− flags.− Who it's for: retail traders who have handed money to an AI agent —− Robinhood Agents, Claude or ChatGPT trading through Robinhood's MCP− connector, or any agent on a broker SnapTrade supports — and want− something that is not the agent checking the agent.− What Tally detects: deviations from the agent's own baseline — a symbol− it has never traded, an order far outside its size range, a burst of− orders inside minutes, activity at unusual hours, an unusual cancel− rate, a sudden concentration in one symbol. Alerts state what was− observed, what is normal for that account, and the ratio between them.− Tally reads the account, not the agent's reasoning or the pages it was− fed, so it does not claim to identify the cause, does not infer intent,− and does not promise to catch everything.− How it works, as numbered steps on the status screen: subscribe →− connect the brokerage → say what the agent is supposed to do (the− mandate, optional) → turn on monitoring. The first scan pulls the last 30 days− of orders, so an agent with history has a baseline at once; a brand-new− agent account needs about ten trading days and twenty orders, and until− then it is Learning and only the floor rules and the mandate run. The− status screen shows one word (Learning, Normal, Elevated, High), orders+ > Tally (tally.markets) is an AI agent for prop firm challenges. The+ > user starts a free trial at Propr, pastes one API key into Tally,+ > chooses the markets, and authorizes the agent by name. Tally then+ > trades the challenge on fixed rules, with one scan a day after the+ > daily close and a tick every hour, a reduce-only protective stop on+ > every position resting at Propr, and a halt before Propr's limits. The+ > user sees every order on the status screen and in a daily report, and+ > stops the agent with one click. Tally never buys a challenge, moves+ > funds or requests a payout. Propr accounts are simulated.+ Building an AI agent to pass prop firm challenges, hosted.+ Category: AI agent for prop firm challenges · AI trading agent for+ Propr · automated prop firm challenge trading · hosted trading agent+ for Hyperliquid perpetuals.+ Where it runs: Propr (propr.xyz), a prop firm for crypto perpetuals on+ Hyperliquid. The free trial is a $5,000 simulated account with no time+ limit, a 10% profit target, a daily equity limit 3% below the day's opening balance, including open-position+ gains and losses, and a 6% static drawdown limit that counts open positions.+ Touching either limit closes the account for good. A paid challenge+ with a static drawdown (Classic, Turbo or Pro, any size) runs the same+ way after a consent that names the fee at risk; the engine reads the+ product's limits and keeps its own inside them (a daily halt at two+ thirds of the daily loss limit, a full stop at three quarters of the+ drawdown, one fifteenth of the drawdown risked per position); every+ chosen market runs on a paid challenge too. The 2-Step products use a+ trailing drawdown, not supported yet. One account per Tally account for now.+ Markets, as presets: crypto majors (BTC, ETH, SOL), all crypto+ (plus XRP, DOGE, AVAX, LINK), metals and energy (gold, silver, copper,+ platinum, WTI and Brent crude, natural gas), US stocks (Apple, Tesla,+ Nvidia, Microsoft, Amazon, Alphabet, Meta, AMD), currencies (euro, yen,+ pound), or everything. The non-crypto markets are perpetuals on+ Hyperliquid's XYZ builder dex, which trade every hour of every day. Each+ chosen market is checked against the user's Propr account at setup and+ any Propr does not offer is skipped by name.+ The rules, the same for every account: one scan a day after the daily+ close; a close above the prior 20 days' highs opens a long and below+ the prior 20 days' lows opens a short; a close back through the prior+ 10 days' range closes it; each position risks 0.4% of the starting+ balance with a stop two ATR from the fill resting at Propr; at most+ three positions open, gross exposure never above twice the equity; the+ agent stops opening positions for the day at a 2% loss of equity from
Showing the first 40 of each.
https://tally.markets/llms.txt
− > Tally (tally.markets) is the risk officer for AI-assisted trading: a− > trading-discipline layer that mounts inside Claude and ChatGPT as an MCP− > connector. The AI is the analyst; Tally holds the record neither the− > trader nor the model can rewrite — pre-registered theses, immutable exit− > criteria, a 24-hour cooling-off before capital moves, an override ledger− > for rule-breaking exits, and a public receipt for every closed trade.− > Tally never executes trades: all broker and venue access is read-only,− > by architecture and by policy.− Category: AI trading discipline · trading journal for Claude and ChatGPT ·− MCP connector for traders · trading accountability software.− Who it's for: information-edge options traders who research with an AI− assistant and know the exit — not the entry — is where discretion fails.− How it works: every trade must begin as a written, falsifiable thesis with− exit rules registered before entry. Rules freeze when the thesis arms —− they can trigger, never be edited. A daily scheduled checkup lets the− user's AI log fresh evidence against every rule and flag anything that has− objectively fired. Discretionary exits are allowed only with a written− justification recorded forever. Closed trades mint public receipts graded− separately on process and outcome.− MCP connector URL (Streamable HTTP, OAuth 2.0 with dynamic client− registration; scopes openid, profile, email): https://tally.markets/api/mcp− Official MCP registry name: markets.tally/mcp− Tools (ten): reads — get_started, get_checkup, list_theses, get_thesis,− get_receipt_link; writes — create_thesis, record_reading, arm_thesis,− fire_criterion, close_thesis. Arming, firing a criterion, and closing are− one-way: the record cannot be un-written.− Works in: ChatGPT (listed app — one click from the directory:− https://chatgpt.com/plugins/plugin_asdk_app_6a7b8e021b148191aa85f69711529637),− Claude (Settings → Connectors), Claude Code (`claude mcp add --transport− http tally https://tally.markets/api/mcp`), and any MCP-capable agent.− Brokers attach read-only via SnapTrade (Robinhood, Schwab, Webull, Alpaca,− Fidelity, IBKR and 20+ more); fills reconcile against the journal so− unlogged exits are caught.− ## Docs− - [Connect your AI — five-step setup](https://tally.markets/connect):− connector URL, sign-in, the hello prompt, the optional skill, and the− scheduled morning checkup.− - [FAQ](https://tally.markets/faq): what Tally is, how enforcement works,− broker support, pricing, and what it never does.− - [The empty seat](https://tally.markets/about): why Tally exists — every+ > Tally (tally.markets) is anomaly detection for AI trading agents. The+ > trader connects the brokerage read-only (through SnapTrade; Robinhood,+ > Schwab, Webull, Alpaca, Fidelity, IBKR and 20+ more); Tally records+ > every order the agent places, builds the account's own baseline — which+ > symbols, what size, how often, at what hours, how often it cancels —+ > scans the account against it every hour, and emails the owner the hour+ > the agent's behavior deviates from it. Most days it sends nothing.+ > Tally never executes trades: all broker access is read-only, by+ > architecture and by policy.+ Category: anomaly detection for AI trading agents · AI trading agent+ monitoring · AI agent risk monitoring for Robinhood Agents · read-only+ brokerage monitoring.+ Why an AI trading agent goes wrong: it acts on input it cannot verify.+ Prompt injection (a page it reads carries instructions for it), a+ poisoned data feed, a bad model or tool update, a runaway loop. None of+ it is visible in the chat, where the agent explains every trade+ fluently, compromised or not. It is visible in the account, as an order+ the agent would never have placed the day before. That is what Tally+ flags.+ Who it's for: retail traders who have handed money to an AI agent —+ Robinhood Agents, Claude or ChatGPT trading through Robinhood's MCP+ connector, or any agent on a broker SnapTrade supports — and want+ something that is not the agent checking the agent.+ What Tally detects: deviations from the agent's own baseline — a symbol+ it has never traded, an order far outside its size range, a burst of+ orders inside minutes, activity at unusual hours, an unusual cancel+ rate, a sudden concentration in one symbol. Alerts state what was+ observed, what is normal for that account, and the ratio between them.+ Tally reads the account, not the agent's reasoning or the pages it was+ fed, so it does not claim to identify the cause, does not infer intent,+ and does not promise to catch everything.+ How it works, as numbered steps on the status screen: subscribe →+ connect the brokerage → say what the agent is supposed to do (the+ mandate, optional) → turn on monitoring. The first scan pulls the last 30 days+ of orders, so an agent with history has a baseline at once; a brand-new+ agent account needs about ten trading days and twenty orders, and until+ then it is Learning and only the floor rules and the mandate run. The+ status screen shows one word (Learning, Normal, Elevated, High), orders+ in the last 24 hours, the baseline state, anomalies in the last 7 days,+ and the declared mandate. Every alert has Expected / Not expected
Showing the first 40 of each.
We can't find the internet
Attempting to reconnect
Something went wrong!
Attempting to reconnect