Back to algoagent.in

Risk Disclosure

Algoagent Core — algoagent.in

Last Updated: August 21, 2026

Please Read This Carefully

By downloading, installing, or using Algoagent Core, you acknowledge that you have read, understood, and accept the risks described in this document.

1. Who We Are — and Who We Are Not — and What the Software Does

Algoagent Core is a software product, provided free of charge. We are a software vendor only. We are NOT a broker, dealer, or exchange; a registered investment adviser or portfolio manager; a custodian of your funds or assets; or a party to any trade you execute. We do not manage funds, do not hold client money, do not execute orders on your behalf, and do not provide personalized investment, financial, legal, or tax advice. We owe you no fiduciary duty. The Software is a research tool — it never provides advice, research content sold as such, or market predictions.

Algoagent Core is a research terminal. It never executes anything. It syncs your holdings through a read-only broker connection (Zerodha Kite) or manual CSV import, runs a nine-stage research framework over each holding — or, on the optional Lite Path, six of those nine, with a raised margin-of-safety requirement and a capped confidence score — and produces structured research dossiers. An IPO is researched under a different nine-stage framework of its own. It does not submit, stage, prefill, relay, modify, cancel, or execute any order — there is no order capability in the Software at all. The AI's output is informational research only, not advice.

The trading risks in Sections 2–7 below describe trading financial instruments generally. They apply to any trading you conduct yourself, outside the Software, in your own exchange. Because the Software places no orders and never will, it introduces no execution risk of its own.

Human in the loop — research, not execution. The human-in-the-loop model here is about judgment, not order approval: the AI researches and produces an assessment; you decide whether, what, and when to trade, and you act in your own exchange. Algoagent will never place, modify, or cancel an order on its own — that is a permanent design boundary, not a roadmap item.

2. General Trading Risk Warning

  • Capital is at risk. The value of your investments can go down as well as up, and you may not recover the amount you invested.
  • You can lose money. Trading can result in the loss of part or all of your trading capital, and in some cases losses may exceed your initial deposit.
  • Past performance is not indicative of future results. Historical returns, backtests, and prior outcomes do not guarantee, predict, or imply any future performance.
  • You should only trade with funds you can afford to lose.

3. Artificial Intelligence and Algorithmic Trading Risks

Algoagent Core uses large language models (LLMs) and algorithmic analysis to generate market research and structured research dossiers. You understand and accept that:

  • The AI can be wrong. Outputs may be inaccurate, incomplete, misleading, or based on flawed assumptions. There is no guarantee of accuracy, completeness, or profitability.
  • Hallucination. Generative AI models can produce confident-sounding statements that are factually incorrect or fabricated.
  • Model error and bias. Models may misinterpret data or reflect biases in their training data.
  • Stale or incorrect data. Analysis may rely on delayed, incomplete, or erroneous market data.
  • Overfitting. Strategies that perform well on historical data may perform poorly in live markets.
  • No guarantee of profit. Nothing produced by the Software constitutes a promise, prediction, or assurance of gain.

What the Software does about this, and what it cannot do. The Software constrains the model rather than trusting it: a source can only be cited if the app actually retrieved it and kept a local receipt, figures asserted in a claim are checked against that retrieved text, growth and ratio figures are computed rather than accepted, scores and verdicts are computed from recorded evidence, and missing data is reported as unknown. These reduce a specific and common failure — a real-looking citation attached to a number the source never stated. They are not a truth test, and you should not read them as one:

  • Checking that a number appears in a document does not establish that the number is right, that the document is right, or that it is the relevant number to be looking at.
  • A claim can be perfectly grounded in its source and still be misleading in context, out of date, or beside the point.
  • The narrative around a checked number — the reasoning, the emphasis, the conclusion drawn — is still model-generated and can still be wrong.
  • Computed figures inherit every error, omission, and later restatement in the underlying data.
  • Confidence is a computed heuristic, not a probability. It reflects how much of the framework was filled in, how much evidence held up when checked, how old that evidence was, and how primary the sources were. A high confidence value is not a claim that the assessment is correct.

You are responsible for independently verifying any suggestion before acting on it. Do not treat AI output as financial advice.

4. Equity and Market Risks

Trading equities involves market risk, company-specific risk, liquidity and trading-hour risk, and settlement and corporate-action risk. Prices may decline due to economic conditions, company performance, interest rates, geopolitical events, and investor sentiment. (Algoagent Core syncs equity holdings read-only via Zerodha Kite; these risks apply to any equity trading you conduct in your own exchange.)

5. Leverage and Derivatives Risk

If you trade leveraged products, margin, futures, options, or other derivatives: leverage magnifies both gains and losses; losses may exceed your deposited funds; and derivatives are complex and not suitable for all investors.

6. Technology and Operational Risks

Technology-related risks include software bugs and defects; connectivity and latency failures; API and third-party outages (brokers and exchanges may experience downtime or rate limits outside our control); your machine and operating system (performance depends on your hardware and configuration); and data accuracy (market data may be delayed or incomplete). You are responsible for maintaining a secure, up-to-date environment; for reviewing research output and verifying it against independent sources; and — for any trading you conduct in your own exchange — for confirming that orders are submitted and filled as intended.

7. Broker Keys — Read-Only Only

Algoagent Core connects to your broker (Zerodha Kite) with your own keys, used solely to sync your holdings. It only ever calls read-only endpoints, and order paths are blocked at the network layer — so even a key that could trade elsewhere can never place an order here. You are solely responsible for creating, securing, and managing those keys and the permissions you grant them — use read-only scopes and disable withdrawals; your relationship with your broker is governed by that provider's own terms; and we never custody your funds or keys. Treat any API keys as sensitive secrets.

8. Research Output Is Not a Guarantee

Algoagent Core produces research dossiers with computed ratings — quality scores, valuation verdicts, scenarios, and confidence levels. These are assessments derived from data, not predictions. A favorable verdict does not guarantee a favorable outcome; scenarios may not play out as described; and the underlying data may be stale, incomplete, or wrong. Any decision you make based on research output — and any trade you execute in your own exchange — is at your own risk.

9. Data Coverage, Entitlement, and Abstention

The Software researches with the data your own keys can reach. That is a real limit, and it is yours to weigh.

  • A key does not tell you what it is entitled to. Provider plans differ, and a key that works for one purpose may be refused for another — a free broker tier can sync your holdings perfectly and refuse every market-data request. The Software probes each provider and reports what it can and cannot get, but it cannot obtain data you are not entitled to.
  • A dossier may be incomplete, and will say so. Where a source is unavailable, a field comes back unknown and a stage may abstain entirely. This is deliberate — the alternative is a plausible-looking number nobody can check — but an incomplete assessment is still an incomplete assessment, and the missing pieces may be the ones that mattered.
  • You may be asked to supply data yourself. Anything you supply is only as good as you made it, is labelled as yours, and is weighted below a first-party filing. A dossier built on it is marked accordingly.
  • Coverage varies by market and by holding. Some instruments, and some kinds of data — including on-chain metrics for digital assets — have no source the Software can reach at all. For those, it abstains rather than estimates.

10. Staleness, Revisions, and Withheld Assessments

  • Every assessment is dated. A dossier reflects the data available when it was produced and nothing after it. The Software marks a dossier stale as its evidence ages; it does not retroactively update an old assessment.
  • A detected red-flag event withholds an assessment; it does not replace it. Where the Software detects a significant adverse event affecting a holding — an auditor resignation, a regulatory action, a rating downgrade, and similar — it withholds the affected assessment behind a notice and invites you to re-run the research. It deliberately does not adjust the score. Treat a withheld assessment as "this may no longer be reliable", never as a new conclusion.
  • Detection is not judgment, and absence of detection is not safety. These detectors match published items deterministically. They can miss an event, match one that turns out to be wrong or immaterial, and cannot tell you whether an allegation is true. A dossier that has not been frozen is not thereby endorsed.
  • Re-running research produces a new revision, not a correction of the old one. Earlier revisions remain readable so you can see what changed. A later revision is not proof that the earlier one was wrong, or that the later one is right.

11. News Is Context, Not a Live Feed

The Software gathers news to give research current context. It is not a market-data feed and must not be relied on as one.

  • It is not real time. Items arrive through public feeds and APIs with delays that vary by publisher and are outside our control. The Software labels each item's age and marks stale ones rather than implying everything is current.
  • It is not complete. Coverage depends on which sources publish, which are reachable, and which optional providers you have configured. Absence of news about a holding is not evidence that nothing happened.
  • Publisher errors pass through. Headlines can be wrong, premature, or later retracted. The Software records the source and timestamp so you can check the original; it does not verify the claim.
  • Availability is not guaranteed. Feeds and APIs can fail, rate-limit, or change without notice. The Software degrades to reporting the gap.

12. Not Investment Advice

Nothing in Algoagent Core — including any AI-generated analysis, suggestion, signal, or content — constitutes investment, financial, legal, or tax advice, or a recommendation to buy, sell, or hold any instrument. You are solely responsible for your own investment decisions, your own research, any tax obligations, and compliance with any broker's terms (for trading you conduct in your own exchange) and applicable law.

13. Regulatory and Compliance Context (India)

If you are located in India, you acknowledge that trading in securities and other instruments is subject to the regulations of the Securities and Exchange Board of India (SEBI) and other applicable authorities. You are solely responsible for ensuring that any trading activity you conduct complies with all applicable Indian laws.

Algoagent is a software vendor. We are not registered with SEBI in any capacity — not as an investment adviser, not as a research analyst, not as a portfolio manager, and not as a broker or other intermediary. We do not hold ourselves out as any of these.

We want to be precise about why, because the Software produces documents it calls research dossiers, and those dossiers contain a valuation assessment and a bear/base/bull range:

  • The Software is licensed to you as a tool, not sold to you as research. What you pay for it is nothing.
  • Every dossier is generated on your own Mac, by a model you supply the key for, from data your own credentials retrieved. We do not run the model, do not see the output, and do not keep a copy.
  • We do not publish, distribute, circulate, or sell research reports, to you or to anyone. There is no research we authored, no recommendation list, no model portfolio, and no target price we stand behind.
  • We receive no consideration — from you, from any issuer, from any broker, or from any third party — in connection with any security the Software researches. We hold no position we are talking our own book on.

A dossier is output you generated with a tool you licensed. It is not a research report we issued, and it must not be relied upon as one.

14. No Fiduciary Relationship; Limitation of Liability

No fiduciary, advisory, agency, or partnership relationship is created between you and Algoagent by your use of the Software. Algoagent does not guarantee any trading outcome and is not liable for any trading losses, whether direct, indirect, incidental, consequential, or special. All trading decisions are made by you, and you bear full responsibility for the results. Nothing in this document excludes or limits any liability that cannot lawfully be excluded.

15. Recommended Practices

To help manage your risk, we strongly encourage you to: treat research output as one input to your own judgment, not the decision itself; never risk money you cannot afford to lose; do your own research and independently verify AI suggestions; understand the products you trade; use appropriate risk controls such as position sizing and stop-losses; and keep your keys, machine, and Software secure and up to date.

16. IPOs and Grey Market Premium

If you research an initial public offering with the Software, one stage of that framework records a grey market premium (GMP) snapshot — the price at which an unlisted offering is being quoted in an unofficial, off-exchange market — together with its timestamp and the source it came from.

You should understand what that number is and is not:

  • The grey market is unregulated and unofficial. It is not an exchange, it is not supervised by SEBI, and there is no settlement guarantee, no price discovery mechanism you can audit, and no recourse if a quote is wrong or manufactured.
  • SEBI has repeatedly cautioned retail investors against treating grey market premium as a signal of an offering's value or of its likely listing performance. It is frequently thin, easily manipulated, and a poor predictor.
  • The Software records GMP as sentiment, not evidence. It is flagged excluded in code and cannot reach the valuation stage — the valuation assessment for an IPO is computed without it. It is shown to you because concealing a number the market is talking about is its own kind of dishonesty, not because we think you should act on it.
  • Nothing in a dossier, including a GMP snapshot, is a recommendation to apply for, subscribe to, or avoid any offering.

An IPO carries risks a listed holding does not: no trading history, a prospectus written by the issuer, a promoter who may be selling down, a valuation nobody has tested, and listing-day volatility that can be severe in both directions. Read the offer document. Section 12 applies here in full.

17. Contact

If you have any questions about this Risk Disclosure, please contact us at:

Email: contact@algoagent.in

Bug reports: bugs@algoagent.in

Website: algoagent.in

This Risk Disclosure does not constitute an offer or solicitation to trade any financial instrument. Trading involves risk. Algoagent Core is portfolio research software with no order capability — it never executes anything. Use it responsibly and at your own risk.