AI tool listing.
Nashnova is an AI investment-research workbench built around a useful distinction: an investment thesis should persist after the first question. The current product lets a user start from a company, asset, sector, macro variable or market theme, formulate a view, attach evidence and risk variables, and keep revisiting that view as earnings, prices, filings, news or other signals change.12 That makes the core unit of work closer to a research process than a chat response.
The web product currently combines general financial Q&A, scheduled briefs, an Agent plaza, an “investment avatar” or digital twin, market/news pages and saved research paths.13 The upgraded app extends the same model to iPhone, Android and web: a user can follow questions or securities, create recurring research tasks, subscribe to agents and build a digital twin that learns which assets, risk variables, frameworks and output style the user wants it to monitor.3 Nashnova explicitly says the purpose is to preserve and update a research identity, not to hand control of investment decisions to an autonomous trader.
The data layer is considerably broader than the old “HK & US equities” label suggests. Current documentation describes support for selected equities in the United States, Hong Kong, mainland China, Japan and South Korea, plus ETFs, indexes, foreign exchange, commodities and crypto.45 Depending on the workflow, Nashnova can combine market data, financial statements, filings, earnings releases, conference-call material, global and Chinese news, macro series, research materials, company relationships, unusual options activity, dark-pool data, congressional disclosures, insider transactions and Web3 signals.4
This breadth is useful because investment questions usually cross data silos. A material earnings move may require current price/volume, the release, management language, analyst research, options positioning and macro context. Nashnova's agent runtime is designed to choose relevant financial tools, stream the research process and assemble output that keeps citations and timestamps attached where available.4 The current documentation also describes run tracing, multi-turn context, scheduled agents, model routing and deployable service primitives, suggesting that the consumer research interface sits on top of a more general financial-agent runtime rather than a fixed prompt template.
Research Agents provide different analytical lenses. Nashnova describes frameworks such as fundamental/company quality, supply chains, macro cycles, commodities and contrarian risk.4 A useful disclosure is that agents inspired by a person or institution are built from public materials and do not represent that person's or institution's current views.4 This matters because the app now sells subscriptions to named agents: a stylistic/research-framework simulation should not be mistaken for access to the named analyst or to unpublished current recommendations.
The newer digital-twin feature applies that personalization idea to the user. Users can define assets and themes to watch, recurring tasks and how analysis should be communicated; the app can then publish or share agent answers and push updates.36 This can create a valuable research habit—explicit variables, scheduled review, retained reasoning—but it also increases the amount of user-specific investment context that the service may hold, which raises the importance of clear privacy and deletion controls.
Daily briefs and monitoring are central rather than decorative. The current documentation says briefs can track securities, variables and frequencies, while scheduled workflows support pre-market, post-market and thematic monitoring.4 App Store users specifically praise position/thesis monitoring and daily reports as the reason they return to the product.6 That qualitative evidence fits the intended workflow better than a generic “AI answer quality” score would.
The product also includes an AI paper portfolio and investment journal in its mobile store description.67 These are appropriate educational/review tools because they let users compare a decision process with later outcomes without turning Nashnova into a broker. The service's formal public boundary is clear: it does not execute trades, hold client assets or give an agent authority to buy or sell on a user's behalf.8
Source traceability is a stated design goal. The methodology says useful output should show the source/citation where available, data or publication time, assumptions, material risks/counterarguments and a practical checkpoint for later review.2 The data policy similarly advises users to verify original filings, announcements, transcripts and datasets for material decisions.5 This is the right posture for financial AI: citations help audit the evidence trail, but do not guarantee that the model interpreted the source correctly.
Nashnova also exposes meaningful limits instead of making a universal real-time claim. Coverage and freshness depend on market, license, region and workflow; some feeds update intraday, while filings, transcripts and research follow their own publication/ingestion schedules.54 The company does not publish one universal latency SLA for all data. That is less marketable than “real time everywhere,” but more credible.
The principal product risk is the same breadth that makes it useful. An answer can combine multiple licensed/public sources, model routing, derived metrics and agent interpretation. If a source is delayed, corrected, differently adjusted or unavailable in a region, the generated thesis may inherit that issue.5 Nashnova's own methodology warns that AI output can be incomplete, inaccurate, delayed or internally inconsistent.2 Users should therefore treat the platform as a research accelerator and record-keeping system, not an oracle.
Nashnova now publishes current web pricing clearly. The Free web plan costs $0 and currently includes no monthly Credit grant.9 Pro costs $20 per month with 20,000 Credits per month; Elite costs $100 per month with 100,000 Credits; Ultra costs $200 per month with 200,000 Credits.9 The linear relationship between plan price and included Credits makes the basic subscription economics unusually easy to understand.
One-time top-up packs are also public: 8,000 Credits for $10, 27,000 for $30 and 50,000 for $50, with actual purchase eligibility confirmed in the signed-in flow.9 Apple exposes the same three Credit packs and the same Pro/Elite/Ultra subscription prices as in-app purchases, which provides an independent storefront cross-check of the current public web price structure.6
Apple additionally lists separate $20/month purchases for named research agents such as Liu Yuhui, Hong Hao and Fu Peng.6 These should be understood as optional agent subscriptions, not as inclusion in the ordinary Pro plan unless the live checkout explicitly says otherwise. The August 2026 App Store release notes confirm that agent subscriptions are a newly introduced feature.6
The free mobile apps can be downloaded without purchasing a subscription. The current App Store lists the app as Free with in-app purchases; Google Play also lists in-app purchases and more than 1,000 downloads.67 Nashnova's web documentation says public material can be browsed without signing in, while starting a conversation may require authentication.4
Distribution is now broad. Nashnova provides iPhone/iPad, Google Play, an Android APK and a web app.3 This is materially different from an early beta-only product. The direct APK route is convenient in markets where Google Play is inaccessible, but users should prefer the platform store when available because the direct APK is not reviewed through Google Play's distribution process.3
The pricing page says recurring subscriptions can be cancelled through the original subscription channel and remain active through the paid billing period. Virtual services are generally non-refundable after sale except where applicable law requires otherwise.9 Prices, Credits and eligibility can change, with checkout controlling the actual transaction terms.
The main unresolved pricing question is what a Credit buys across different research operations. The public pricing page makes plan quantities clear, but the reviewed page does not provide a simple endpoint-by-endpoint consumption table comparable to developer APIs that define exactly how many credits each tool call costs.9 A user can compare subscription price and included volume, but predicting monthly consumption for heavy multi-agent research still requires observing the live product.
Value therefore depends on workflow density. At $20/month, Pro is competitively positioned if scheduled research, source synthesis and thesis tracking replace several hours of manual work. Elite and Ultra become harder to judge without a published normalized workload model: the Credit quantity scales proportionately with price, so the justification must come from actual research volume rather than a steep unit discount.9
Nashnova has a better external evidence base than many new agent products, but it is still small. Current independent coverage is concentrated in two exact-product store surfaces—Apple App Store and Google Play. There is no current G2/Capterra/TrustRadius-style enterprise review corpus and no independent exact-product benchmark in the broader review evidence.
The strongest current user signal is Apple. At the latest readback on August 28, the US App Store showed 4.9/5 from 12 ratings.6 Reviews repeatedly praise the calm interface, daily/portfolio briefs, thesis tracking, automated monitoring and the way answers present evidence and risks rather than only a conclusion. Some reviewers also identify early-product roughness, including interaction polish and dark-mode refinement.6 The sample is exact-product and current, so it is a usable quantitative user signal, but 12 ratings is far too small to carry the weight of a mature consumer-app corpus.
Google Play provides a second distribution/reliability signal rather than quantitative user-rating evidence. The current listing shows 1K+ downloads, in-app purchases and an August 25, 2026 update.7 Its developer-declared Data Safety section says personal information, app activity and device/other identifiers may be collected; app/performance information and identifiers may be shared; data is encrypted in transit; and deletion can be requested.7 No numeric Google Play rating was visible in the captured store result.
The App Store privacy disclosure is narrower: it says email/contact information and user ID may be linked to identity, while usage/diagnostic data may be collected without being linked.6 Apple explicitly says this information is developer-declared and not verified by Apple. The mismatch is not necessarily contradictory because Apple and Google use different disclosure taxonomies and app implementations can differ, but the user should rely on the underlying privacy policy for the full processing picture.
That is currently a documentation weakness in the current evidence review. Apple links to a Nashnova/CMCF privacy policy and the product's own methodology references a user agreement and privacy policy, but the underlying policy page did not render substantive text in the reviewed web capture.62 The reviewed evidence does not support an independent summary of current retention periods, AI-training use, model-provider transfers, international-transfer terms or detailed deletion mechanics from the legal policy itself. For a product that can learn a user's holdings, themes and research habits, those details should be easy to audit publicly.
Nashnova's public investment boundary is much stronger. Its dedicated disclaimer says the product provides information, market intelligence and research support; it is not a licensed investment adviser, broker or financial planner; it does not execute trades or hold assets; and AI Agents cannot decide on the user's behalf.8 It also explicitly warns that AI can omit context, misread a source, use stale information or present an inference too confidently. That is unusually direct and appropriate for the category.
Data-source transparency is mixed in a more defensible way. Nashnova lists the classes of data it can use and explains coverage, timestamps, citations and limitations, but does not publicly enumerate commercial data suppliers or research-provider relationships.54 The company says this is partly because provider disclosure depends on authorization and must stay current. That is reasonable commercially, but it means an external reviewer cannot verify source-by-source licensing, latency or completeness from the public site.
Copyright/access-control positioning is explicit. Current documentation says the platform uses public information, authorized content or compliant interfaces and, for subscription-restricted/copyrighted material, limits processing to permitted metadata, structured fields, compliant summaries, necessary short quotations and links rather than reproducing protected works or bypassing access controls.4 This is a positive governance claim, though The reviewed evidence does not include an independent audit of each underlying data agreement.
Nashnova has also published a first-party comparative test against Claude with web search and GPT with web search. The company says it ran the same 100 prompts across the systems, de-identified/randomized outputs and blind-scored five categories; it reports Nashnova leading overall at 79.2 and winning 95 of 100 questions, versus 64.4 for Claude and 55.9 for GPT.10 The methodology is more concrete than a vague “better than ChatGPT” claim, and the claimed categories—grounding, source traceability, factual precision, reasoning, actionability, completeness and timeliness—are relevant to investment research.
It is still a vendor-run benchmark. The reviewed publication does not establish that scoring was commissioned or reproduced by an independent laboratory, and Nashnova designed both the product and the comparison.10 The result is useful evidence that Nashnova has attempted a structured evaluation and a hypothesis about where the product should outperform general assistants; the 95/100 wins and 79.2/100 figures are not treated as external rating inputs.
The benchmark also cannot prove live market reliability. Financial research quality changes with the exact securities, source licensing, market regime, publication timing and model version. Nashnova's own data policy correctly says coverage varies and asks users to verify material claims at original sources.5 A more mature evidence base would include repeatable external tests across stale-data traps, earnings restatements, conflicting research, source attribution, historical point-in-time queries and non-US coverage.
The mobile review population is encouraging but young. Apple's version history shows rapid iteration from June through late August, with features such as agents, brief templates, tracking tasks, push notifications and publishing changing within weeks.6 Fast shipping is positive, but also means older user impressions can describe a materially different product. The present evidence supports medium confidence rather than high confidence.
Nashnova is best suited to self-directed investors who already think in theses but struggle to maintain the process. A user can write why a company is attractive, define the variables that would invalidate the view, receive updates when evidence changes and later review whether the original reasoning survived contact with reality.26 That is a much healthier use of AI than asking “what stock should I buy?” and treating a fluent answer as advice.
It is particularly strong for investors who operate across US and Greater China markets. Current documentation explicitly covers US, Hong Kong and mainland China and incorporates multilingual news/research sources, while also extending into Japan, South Korea and cross-asset data where supported.54 Investors whose edge depends on comparing Chinese and English-language information may find that combination more useful than a US-centric general assistant.
Earnings-focused users are another natural fit. Nashnova can combine structured financials, earnings-call materials, research views, news and price reaction into an ongoing question about what changed and what should be monitored next quarter.4 The product's methodology encourages users to separate facts from interpretation and identify failure conditions rather than merely summarize the release.2
Users who follow institutional/public-flow signals can also benefit from congressional transactions, insider disclosures, unusual options and dark-pool data as supporting context.4 Nashnova explicitly warns that such signals do not by themselves establish intent, which is important: a large option trade or insider sale should be a research lead, not an automatic buy/sell trigger.
The Agent plaza and digital twin are most useful for users who want recurring analytical lenses. One agent can pressure-test a company fundamentally while another examines macro or supply-chain variables; a user's own avatar can preserve personal questions and monitoring style.34 Users must remember that named/inspired agents are simulations based on public material, not live representation by the person or institution.4
Nashnova is less suitable for anyone seeking automated execution, fiduciary advice or a regulated adviser relationship. The service explicitly provides none of those things.8 A user who needs suitability analysis, portfolio construction tied to personal legal/tax circumstances or actual order execution needs a different service and, where appropriate, a licensed professional.
It is also a weaker fit for high-frequency traders who require a contractual universal real-time feed. Nashnova says latency varies by data type and does not publish one universal production-latency promise.5 Source-linked research and scheduled monitoring are different products from a co-located trading feed.
Institutional buyers with strict data-vendor procurement requirements may want more disclosure than the public site offers. The documentation describes categories and compliance approach but does not enumerate specific commercial data suppliers.5 That may be resolvable in a commercial diligence process, but it cannot be verified from the public review surface.
Privacy-sensitive users who intend to load detailed holdings, strategies or recurring investment questions should review the live user agreement/privacy policy before doing so. Store disclosures provide useful high-level data categories, but the legal privacy page was not substantively auditable in this review.67
Finally, users who mainly want occasional generic finance explanations may not need a specialized product. A general-purpose assistant can explain concepts and summarize public web sources. Nashnova's value appears when the work becomes persistent: source-linked thesis formation, scheduled monitoring, multiple financial data tools and review of how a judgment changes over time.102
Products matched by shared tasks, audience, workflow, product type and capabilities.