How can we help?

Satcove is a multi-AI consensus engine for decisions. When you ask an important question, it can query 6 AI models simultaneously (Claude, ChatGPT, Gemini, Mistral, Perplexity, Grok) and synthesize their responses into a structured answer with agreements, divergences and an agreement score.

ChatGPT and Claude usually answer from one provider at a time. Satcove compares six independent providers in parallel, names material conflicts, and suggests what to verify. Several models can still share the same blind spot, so agreement is not proof of accuracy.

You can run 1 consensus trial without an account. A free account includes 3 consensus runs per week, with no credit card required.

Yes. If you write "consult all AIs" or "ask all models" (or in French: "demande à toutes les IA", "consulte tous les modèles"), Satcove will always run a full 6-model consensus regardless of what you're asking. You can also manually trigger it from the consensus button in the interface.

Use Satcove when the answer affects a decision: buying something, choosing a technical option, comparing offers, preparing a client note, or checking whether several models see the same risks. For casual questions, a single model is often enough.

Ask the same question with the same context, then compare the conclusions, assumptions, sources and points of disagreement. Satcove sends one question to several models and synthesizes the result, so you do not need to switch between apps or reconcile six tabs by hand.

Yes. Start a Consensus query to send the same question to several models in parallel. Each model answers independently, then Satcove shows the shared position, the meaningful differences and the next point to verify.

By default Cove answers fast with a single model — useful for everyday questions and follow-ups. For an important question, start Consensus: six providers answer independently, Satcove shows where their conclusions align or diverge, and the synthesis recommends a concrete next step. Agreement is not factual verification.

The agreement score measures how closely the 6 AI answers aligned in conclusion and direction. It does not measure factual accuracy or the probability that the verdict is true. High agreement means the panel converged; low agreement means the models reached different positions and Satcove will show where. Check sources, freshness, and qualified expertise separately when the stakes are high.

Look at the divergences, the assumptions, the freshness of the information, and whether a primary source or qualified expert is needed. Satcove helps expose agreement and disagreement; it does not replace professional advice or primary evidence.

Do not choose the most confident-sounding answer. Isolate the disputed claim, inspect the sources and check the current primary evidence. In Satcove, disagreement is a signal that tells you where to slow down and verify.

No. Agreement measures how closely the models align, not whether they are correct. Several models can repeat the same outdated or incorrect assumption, so important claims still need current sources or qualified human review.

Claude (Anthropic) — strong at reasoning and nuance. ChatGPT (OpenAI) — general-purpose, breadth. Gemini (Google) — multimodal, research. Mistral — European, multilingual. Perplexity — live web search, up-to-date data. Grok (xAI) — real-time internet, sharp reasoning. Each brings a distinct perspective.

Typically 8–15 seconds. You see a live timeline as each model responds in real time. The synthesis is generated after all models have replied.

Board mode decomposes a complex project across 5 specialist roles: strategy, finance, technology, operations and research. It then integrates their contributions into one plan while preserving cross-role challenges.

State the decision, the options, your constraints, and what would make a wrong answer costly. Satcove can compare the models' reasoning, surface disagreements, and turn the output into a more defensible next step.

Give the stack, constraints, traffic, budget, team experience and failure risks. Ask Satcove to compare options across maintainability, security, cost, reversibility and operational complexity.

Today, use Deep Analysis and exports to create a structured decision note. The roadmap is to make Decision Briefs more explicit: conclusion, assumptions, risks, disagreements and next checks in one shareable artifact.

Use one shared workflow for the question and its context instead of copying the same prompt into separate chats. Satcove lets several models work from the same question, then shows their differences in one result.

Cove Fight puts two AI models on opposite sides of a debate. You pick a topic or decision — one AI argues for, one AI argues against. You get the strongest case on both sides, structured round by round. Ideal for stress-testing a business decision, exploring both sides of a dilemma, or simply seeing where the real arguments lie.

Consensus maps convergence and divergence across independent answers, then synthesizes a recommendation. Fight deliberately develops opposing cases before arbitration. Use Consensus for a broad panel view and Fight when the strongest arguments on both sides need to be made explicit. Neither mode turns model agreement into proof of truth.

Yes. Pro users can assign a default role to each AI (e.g. 'corporate lawyer', 'startup founder', 'financial analyst') in Settings → Cove Fight Personas. These personas are injected before every fight so the AIs argue from a specific expert perspective.

Yes. Cove Fight is available in the Satcove iOS app from the main conversation view. Tap the fight icon (⚔️) to start an adversarial debate. Completed fights are saved to your conversation history.

Not currently. A browser-extension build exists in development, but no official Satcove listing is publicly available in the Chrome Web Store. Use the web app, the native iOS app, or the iOS share workflow instead.

Paste or share its URL into Satcove on the web or from iPhone. Satcove compares the available model analyses and identifies claims that still require a primary source.

10 messages per week, 3 consensus queries per week, 7-day conversation history, and access to all 6 AI models. Quotas reset weekly for eligible new free accounts; older accounts may still follow their existing allowance.

Yes. Cancel your subscription in Settings. You'll keep access until the end of your billing period. No cancellation fees.

Refund eligibility depends on the purchase channel and the applicable terms. Contact support@satcove.com with your account email and receipt so we can review your request. Apple subscriptions are handled under Apple's refund process.

iOS subscriptions are managed through Apple. You can manage or cancel them in iPhone Settings → Apple ID → Subscriptions.

Privacy Shield is available on every plan and is controlled by the user. When enabled on a supported text flow, it attempts to replace common identifiers before model calls. Automated detection can miss contextual identifiers, and images, attachments and some tools require separate handling.

Satcove compares multiple AI answers while Privacy Shield helps minimise exposure of identifying details before model calls. Review the privacy policy and the active flow before sharing sensitive information; privacy protection is not a guarantee of accuracy or professional advice.

No. None of the AI providers we use (Anthropic, OpenAI, Google, Mistral, Perplexity, xAI) train on data sent through API calls.

Yes. Go to Settings → Delete Account. This permanently deletes all your data including conversations, memories, and payment history. This action is irreversible.

Yes. Go to Settings → Export my data. You'll receive a JSON file with all your conversations, transactions, and account data.

Yes. Upload PDFs (up to 15 MB) or paste URLs. Satcove will extract the content and analyze it with AI consensus.

Cove Memory lets Satcove remember facts about you across conversations (your preferences, context, previous decisions). You can manage and clear memories in Settings.

Satcove Market lets you scan any product barcode and instantly get a 5-AI product audit: ingredient analysis, health assessment, sustainability, and competitor alternatives. Available on the iOS app.

Satcove Trip provides destination intelligence — ask about any city or country and get consensus-based travel advice from 6 AIs: best time to visit, local tips, safety, budget, and hidden gems.

Satcove currently has 20 in-product language profiles covering 18 distinct base languages. The public App Store listing separately declares 15 localized languages. These counts describe different surfaces.

Satcove takes one question, sends it to several AI models in Consensus mode, compares their conclusions and sources, then presents agreements, disagreements and a practical next step. It is a decision aid, not a guarantee of truth.

Cove selects the appropriate workflow. In Consensus mode, the same question and context are sent independently to the panel, then the responses are checked and synthesized into one structured report.

Six models provide useful perspective diversity while keeping latency, cost and the final report understandable. More models are not automatically better; the useful signal is independent agreement and clearly explained disagreement.

The models answer independently first. Satcove extracts shared conclusions, separates material conflicts, checks whether sources or freshness matter, and writes a synthesis with the next verification step.

Agreement describes how closely the models converge. Synthesis is the readable report that combines their answers, preserves important disagreement and turns it into a practical next step.

Most consensus runs complete in roughly 8 to 15 seconds, depending on the models, source retrieval, attachments and network conditions. The live timeline shows progress while the panel responds.

No. Fresh model responses, source checks and consensus reports require a network connection. You can keep exported information locally, but Satcove cannot run a new multi-model query offline.

Start with the conclusion, then inspect assumptions, sources and divergences. Treat agreement as a signal about model alignment, not proof. For high-stakes decisions, verify primary evidence or consult a qualified professional.

It means the participating models reached similar conclusions for the question and context provided. It can help with orientation, but cannot rule out shared errors, outdated information or missing context.

It means the models reached materially different conclusions or emphasized different risks. Use the disagreement to identify the claim that needs checking instead of choosing the most confident answer.

No score replaces evidence. A high score measures panel alignment, not truth. Check current sources and seek qualified advice when the decision affects health, legal rights or money.

Do not rely on consensus alone for emergencies, diagnosis, legal representation, regulated advice, irreversible transactions or decisions requiring access to private records. Use it to prepare questions and identify what must be verified.

Models can share training-data blind spots, repeat outdated claims, misunderstand context or cite weak sources. Consensus reduces single-model blind spots but does not establish factual certainty.

Depending on the flow and account state, questions and results may be stored so you can continue a conversation or view history. Review the privacy policy and delete history or your account when you no longer need it.

Privacy Shield can replace common identifying details before supported model calls. It is an automated protection layer, not a promise that every contextual identifier will be detected.

Satcove processes the account, usage, billing and content data needed to provide the selected service. The privacy policy describes the categories, purposes and retention periods.

Satcove is designed with European data-protection requirements in mind, including privacy controls and deletion workflows. Consult the current privacy policy and data-processing terms for the applicable details.

Do not assume HIPAA compliance for a general Satcove account. Do not submit protected health information unless an appropriate contractual and technical arrangement has been confirmed with Satcove.

Use Privacy Shield where available, minimise identifying details and avoid uploading information you are not authorised to share. For regulated data, confirm the applicable terms before use.

Conversation-level deletion depends on the current history controls. Use the conversation menu when available; deleting the account removes all account data and cannot be undone.

Pricing depends on the selected plan and purchase channel. The pricing page is the source of truth for current limits, taxes and subscription terms.

Starter is intended for regular individual use. Pro adds higher allowances and professional workflows where available. Check the pricing page for the current plan comparison.

A free entry tier or trial may be available depending on the surface and account eligibility. The signup and pricing pages show the current allowance before payment is requested.

Student discounts are only available when shown in the current pricing or checkout flow. Contact support with proof of eligibility before purchasing if you need confirmation.

Yes. Enterprise discussions can cover access, security, team controls and support requirements. Contact the Satcove team for a current quote and terms.

Yes, you can change plans from the account or billing flow when the option is available. The checkout screen shows the effective date and any prorated amount before confirmation.

The affected workflow is paused or limited until the allowance resets or you change plan. Satcove shows the relevant limit and reset timing when available.

Cove Agent is Satcove's action-oriented assistant for research and workflows. It can ask follow-up questions, preserve context and pause for confirmation before supported external or irreversible actions.

Use the share control on a completed report when available, or export the report and share the resulting file. Remove sensitive details and verify that the recipient is authorised to see it.

Export options depend on the current plan and surface. When export is available, use the report action menu and check the generated file before sharing it.

You can keep reusable prompts in your conversation history or notes. A dedicated favorites control is only available where it is shown in the current product interface.

API access is intended for eligible plans and enterprise workflows. Contact the team for current availability, authentication requirements, limits and data-processing terms.

The default panel is curated for perspective diversity. Custom model selection depends on the current plan and workflow; not every mode supports arbitrary model changes.

Usage and remaining allowances are shown in the account or plan interface when available. Limits can differ by workflow, plan and purchase channel.

Notifications depend on the workflow, device permissions and current app support. Enable Satcove notifications in device settings and use the in-product timeline for the run state.

A query may wait on slow model responses, source retrieval, rate limits or a large attachment. Check the live timeline and retry when the service is available.

Satcove may report a partial panel when a provider is unavailable. Treat it as lower coverage, inspect which model failed and retry later for a complete comparison.

Check your connection, allowance and attachment size. If it persists, record the time, workflow and request ID shown by the app before contacting support.

Use the password-reset link on the sign-in screen. Check spam folders and make sure you are using the account email.

Use the sign-in recovery flow or contact support from the account email. Never send passwords or payment secrets in a support request.

Confirm the email, authentication method, browser cookies and service status. Use the same login method used when the account was created.

Update the app or browser, restart it and retry without the largest attachment. If the crash repeats, send the device, app version, timestamp and steps to reproduce.

Use the Help Center support link or email support@satcove.com with your account email, a short description and non-sensitive diagnostic details.

Support availability and response times depend on the plan and channel. The support team confirms the applicable response window when you contact them.

Check the current service status if available, then retry after a short wait. Include the affected workflow and timestamp when reporting a possible outage.

State the decision, context, constraints, date, location, budget and what the answer must compare. Ask for assumptions, sources and uncertainty instead of a confident guess.

Consensus is most useful for factual comparisons, trade-offs, source checks and decisions with explicit constraints. It is less meaningful for purely subjective or creative prompts.

Yes. Use follow-ups to test assumptions, request sources, isolate a disagreement or ask what evidence would change the conclusion.

Give the research question and scope, ask models to separate facts from inference, inspect citations and record unresolved claims for primary-source verification.

Define options, constraints, reversibility and the cost of being wrong. Ask for the strongest case against the leading option and the checks required before acting.

Share a sanitised report or export where your plan supports it. Keep the decision owner, assumptions, source dates and unresolved disagreements visible to the team.

Use it at review points where an error is costly: before sending a recommendation, committing money, signing a document or relying on a current claim.

Do not confuse agreement with accuracy, omit key context, accept stale sources, hide disagreement or paste sensitive data without checking its handling.

Use Consensus for decisions, Verify for claims and Cove Fight for opposing cases. Compare assumptions and next checks, not only the final verdict.

Keep a reusable question template, require dates and sources, log what changed after verification, and use disagreement to focus human review.