Frequently Asked Questions

exployt is a desktop-first platform for orchestrating AI agents. Instead of being tied to a single vendor, you work with many AI providers side by side and decide how much freedom each agent gets.

Think of it as a spectrum of control. At one end you run simple, predefined routines. At the other, fully autonomous setups where AI agents plan, coordinate, and even manage other AI agents. You stay in charge of the parts that matter to you — tight permissions, fixed workflows, privacy boundaries, approval gates — and hand off the rest.

exployt supports the major AI providers and keeps adding more as the field moves. Out of the box that includes Claude Code, OpenAI Codex, Cursor, Windsurf, and direct API providers such as Anthropic, OpenAI, xAI, Google Gemini, MiniMax, and Qwen. Through OpenRouter you reach 350+ models through a single integration, and through Ollama you can run dozens of local, offline-capable open-source models.

API models aren't just called as plain chat endpoints — exployt wraps them with full agent functionality so they participate in workflows the same way Claude Code or Codex do. You can also switch an agent's model or provider later while keeping its history and context intact.

No. exployt does not send your source code, file contents, local databases, agent history, or project details to our servers during ordinary desktop-app operation. Everything an agent reads, writes, or learns stays in a local database on your machine. Our Privacy Policy spells out exactly which account, authentication, billing, subscription/license, security, fraud-prevention, support/feedback, and consented website analytics data we process for the cloud service — anything outside that scope we simply don't.

Content leaves that local boundary only when you deliberately send it. That includes prompts or context you choose to send to third-party AI providers or tools, Marketplace items you intentionally upload to share or sell, and support or bug reports you submit after reviewing the preview and confirming what will be sent. For Marketplace items that contain code, reusable assets, connector packages, or integration modules, we recommend minimizing the package and using obfuscation where appropriate; exployt can support that workflow.

Partly. Most cloud AI providers need an internet connection, but local models through Ollama run fully offline. The desktop app also has a 3-day grace period for account and subscription checks — after that it needs to reach the exployt server once to revalidate, otherwise subscription-locked features pause until it can.

The desktop app runs on Windows x64 today, and macOS is on the way. Companion apps for iOS and Android let you monitor and steer your agents from your phone. exployt is early-stage and moving fast, so expect more platforms and integrations to land regularly.

No. You can run exployt with a single AI CTO that handles everything — you describe what you want, it plans the work and directs the agents. If you do know how to code, every layer is open to you: agents, workflows, permissions, settings, and debugging tools. The founder of exployt is a self-taught hobby programmer who wrote essentially no code by hand for the product — the agents did. That's the whole point: it should work the same way for you.

Most AI tools start every session from zero. The model forgets your last conversation, the decisions you made, and the bugs you already solved. exployt fixes this by writing everything to a local database on your machine — full chat history, every file change, learnings, decisions, and a dedicated error database. Nothing of this is ever sent to our servers.

That database is searchable not just by keyword but semantically, by meaning, so an agent can ask "what did we decide about authentication?" and get the real answer. Agents store and retrieve learnings, decisions, and wiki-style documentation through their own tools, so context survives across sessions, projects, and team members. The practical effect: an agent picks up exactly where the last one stopped instead of relearning your project every time.

This is the problem we care about most, because debugging is what makes AI genuinely useful instead of just fast at typing. In exployt, an agent in a testing or QA role can do everything a human does in an IDE — and then some.

It can set breakpoints, step through code, inspect real runtime variable values, and walk the live call stack. It can build the app it's developing, start it, stop it, and trigger any method or function inside it. It can even operate the developed app's user interface itself — clicking buttons, filling fields, navigating screens — and capture and read screenshots of the result. So the AI does the tedious QA work, not you.

We also built an automatic debug system: temporary trace logging is woven into the code while an agent works and removed automatically once the task is approved (and stripped from release builds entirely). Agents can pull these traces on demand to understand a whole flow end to end. On top of that, a usage map lets an agent look up which methods call a given method and which it calls — like an IDE's "find usages" — and a structured error database keeps the history of every failure and its fix. Agents can profile too, spotting performance bottlenecks and memory problems rather than just guessing.

Giving feedback is just as easy from your side. With UI Links you can mark any element in the app you're building, and that single click is enough for the AI to know exactly which element you mean — no verbal description needed — along with the history behind it: which agent created or changed it, when, and in which lines of code.

We attack cost from three directions.

First, shadow agents. Any agent can optionally run with a cheap "shadow" model alongside it. The shadow handles everything that isn't demanding — reading into the codebase, finding the right places, preparing materials — and only delegates the genuinely hard parts up to its more expensive main agent. In practice this can cut costs by up to roughly 70%.

Second, model freedom. You're never locked into premium models. Run local open-source models through Ollama, or pick any of the cheaper providers, and assign which work goes to which model.

Third, memory. Because everything an agent learns is persisted and instantly retrievable, agents don't go in circles or re-solve problems they've already cracked. No repeated work means no repeated spend.

You decide which work runs where. Because exployt supports local open-source models through Ollama, you can route sensitive tasks to models that never leave your machine and keep cheaper cloud models for everything else — and switch at the press of a button.

On top of that sits a flexible permission system. A control AI can grant permissions on your behalf, or you can define your own defaults, all the way down to per-role and per-model rules — different agents running different models can have different rights. exployt also replaces the AI models' built-in tools with its own, which is what gives you that finer-grained control. And as covered above, your code and project data stay in your local database; they are never sent to our servers.

Humans are usually the bottleneck — the AI could run around the clock, but it keeps stopping to ask. exployt closes that gap several ways.

Manager agents (AI project managers, or CXO roles like a CTO or CPO) keep the rest of the team moving: they give feedback, make decisions, and pull agents back into the loop when they stall. A decision-management system defines exactly who may decide what — the user, a CPO, a CTO, or a project manager — and which choices must come back to you versus being settled automatically. A status system tells the managers when an agent finishes, has a question, or needs help, so they can step in.

To keep agents busy for hours or days, you queue work with the Targets and Focus systems — a Target can be a whole batch of tasks, features, or goals, defined by you or planned by the AI itself — and Work Modes can instruct an agent to keep going autonomously. And because people are part of the loop too, exployt adds gamification — levels, statistics, an international ranking, and a rewards marketplace — to keep the human in flow so the AI is never waiting.

Agents talk to each other directly over a dedicated agent-to-agent channel (AI2AI). On top of that, manager agents assign tasks, review output, and resolve blockers, so coordination doesn't land on your desk.

You can shape how they collaborate. Build teams of agents, or pair them as coding buddies, and define the rules — who assists whom, who has which permissions, and the hierarchy between them. For complex, parallel collaboration there's a visual, node-based workflow system (comparable to n8n) where you lay out the whole flow up front: which models are used, the permissions, and the exact rules of cooperation. Press start, and the agents work it out among themselves.

Everything is recorded to your local database from the project's very first minute, and everything is visible. In the chat you can toggle each layer independently — tool calls, the AI's thinking, file changes, permissions, and the messages agents send each other — and review all of it, rewound all the way back to the start of the project.

There's a dedicated search across everything, including semantic search by meaning, so you can find any change, decision, or message in seconds. The full file-change history shows precisely which file was changed, when, and by which agent. Crucially that history isn't just for you: when another agent takes over a task, it can read exactly what happened before and continue from there. We already provide token-usage and cost transparency basics, and we're pushing that further.

Both — and that's deliberate. Underneath is a deep, highly configurable system with an enormous range of settings, workflows, and rules. But you almost never have to touch it by hand, because the AI configures it for you. You say "I need a new workflow" or "set up the agents to collaborate like this," and the agents build the workflows and dial in the settings themselves.

If you want the simplest possible experience, you can talk to a single point of contact — a CTO or CPO that even comes as a 3D digital human you can converse with freely, like a real colleague. It takes your requirements and delegates, creates, and manages everything downstream. And a community marketplace lets people share or sell ready-made setting sets and workflows, so you can benefit from others' fine-tuning. Simple when you want it, deep when you need it.

Right — different providers shine at different work, so locking yourself to one is a mistake. exployt lets you route each task to the model that suits it, or pin a specific provider to an agent role or workstream when you want consistency. You can mix premium, cheap, and local models in the same project, and reassign an agent's model later without losing its history or context.

It's usually the opposite. exployt ships with templates, skill-equipped agents, build scripts, and project scaffolding, so onboarding that would take days becomes minutes. Templates come with placeholders and credential slots for services like Microsoft Azure, so you can deploy a website or a classic SaaS service quickly. Many more templates are planned — including community-contributed ones — so the head start keeps growing.

No. The AI ships software around the clock, and you shouldn't have to babysit it from a chair. Companion apps for iOS and Android let you monitor progress, approve changes, and step in from anywhere — the software keeps earning while you're away from the desk.

Click 'Get Started' on the homepage, enter your email, and choose a password. We'll send a verification email — confirm it and you're in.

On the login page, click 'Forgot Password' and enter your email. We'll send a reset link. For security, the reset token is valid for one hour.

Go to Account > Settings > Security and click 'Enable 2FA'. Scan the code with an authenticator app, enter the 6-digit code it shows, and you're set.

Yes. Go to Account > Settings > Profile, enter the new address in the Email field, and click Save. The new address is stored as unverified and we send a verification email — confirm it to finish the change.

Go to Account > Settings > Danger Zone and click 'Delete Account'. We send a confirmation email first. Deletion is permanent and removes the associated account data.

Payments run through Stripe, our payment processor, which shows the available methods at checkout based on your country, currency, and device. That typically includes major cards (Visa, Mastercard, American Express, Discover, JCB, UnionPay) plus PayPal, Apple Pay, Google Pay, Link, bank transfers, iDEAL, Bancontact, BLIK, Przelewy24, SEPA Direct Debit, WeChat Pay, and Alipay where supported. The exact list is shown at checkout for your region.

Yes. Go to Account > Invoices to enter your billing details — company name, address, VAT ID — and download invoices once payments have been created.

Go to Account > Subscription > Cancel Subscription. Your access continues until the end of the current billing period, and you can resubscribe whenever you like.

Refunds are handled under the rules in our Terms of Service. Payments are processed by Stripe; to request a refund or payment support, contact us through the channels listed on our website. Availability and timing depend on the payment method and the applicable processing rules.

Sometimes — and there are two different things here. Free trials: a trial isn't guaranteed; current availability, length, and conditions show up on the Pricing page or at checkout when one is active. Some trials ask for a payment method (such as a credit card) so we can verify each person and stop the same user from claiming a free offer repeatedly; you are not charged during the trial, and if a paid plan is attached it starts automatically when the trial ends unless you cancel first. Free Commercial: separately, if you genuinely can't afford commercial use right now, we sometimes grant free commercial access on a case-by-case basis. We hold no legal claim over what you do with it; we simply ask that, once you're earning enough to easily afford it, you pay what you would have owed — not because we can force it, but because it's fair. See the Free Commercial tab on the Pricing page.

For individuals, yes — that's a core promise. Keeping exployt within reach of a single person on a modest budget is part of the mission, so individual pricing is meant to stay manageable over the long term. Pricing for companies is a different story: business plans are priced higher and we expect business pricing to rise over time, partly to protect the competitiveness of individuals against large organizations.

Yes. Every user gets a personal affiliate link (and promo code) to share. When someone you refer subscribes, you earn 30% of one month of their plan at its standard monthly price — not 30% of an annual or lifetime payment. (Higher Company and Enterprise plans have a higher monthly price, so the reward is correspondingly higher.) The bonus is credited to you once the refund window on that payment has passed — and reversed if they later refund — so payouts are real, not provisional.

How does it work?

Yes. exployt has a community marketplace where you can share — or sell — the setting sets, workflows, templates, plugins, connector packages, integration modules, and other reusable assets that you or your agents create. Uploading a Marketplace item is a deliberate publishing action: the files and metadata you choose for that item are transmitted so they can be reviewed, hosted, shared, sold, or delivered to buyers.

If an item contains code or project assets, you decide what to include. We recommend minimizing the package, removing secrets and private customer data, and using obfuscation where appropriate; exployt is built to support that workflow.

For paid items on the marketplace, creators keep 70% of gross sales and the platform retains a 30% commission, as described in our Terms of Service. The same terms cover eligibility, payouts, and refund handling.

Workstreams are logical groupings for your agents. You might keep a 'Backend' workstream and a 'Frontend' workstream, each with its own agents focused on a different part of the project.

Through AI2AI messages — a dedicated agent-to-agent channel. They also share structured project context such as tasks, todos, learnings, decisions, chat history, file-change history, and lock state, so each agent understands what happened before and doesn't overwrite another's work.

It's the local database in your project that holds all agent activity, chat history, and project metadata. It's safe to add to .gitignore.

Yes. exployt fits around your current setup. It integrates Claude Code, OpenAI Codex, Cursor, Windsurf, and other provider- or API-based agents, so those tools run inside the exployt workflow instead of as separate, disconnected sessions.

Use the Feedback button in the lower-left sidebar inside the app and pick the bug-report option. On the website, signed-in users can open the Feedback page to send bug reports, feature requests, or general feedback. When a support or bug-report flow can include diagnostic details, you see what will be sent before submission and must confirm it; only the data you approve is sent.

Traffic between the app, the website, and our servers uses HTTPS/TLS. Passwords are never stored in plaintext; they are hashed with bcrypt. Selected sensitive server-side fields used for abuse prevention are encrypted with AES-GCM and hashed for lookup. On your machine, local secrets such as cached entitlement tokens, provider credentials, and build/deploy credentials are protected with Windows DPAPI under the current Windows user account. Local project databases and agent history stay on your machine and are not sent to exployt cloud during ordinary desktop-app operation. You can enable local database encryption in the desktop app's encryption settings; when you do not enable it, the local SQLite database remains under your own local device and file-system security.

No. exployt's cloud does not store your AI provider API keys for ordinary desktop-app operation. You enter provider keys in the exployt desktop app's local settings or credential areas, and exployt uses them from your machine for the provider connection you configure. Credential-vault and build/deploy secret stores protect persisted local secrets with Windows DPAPI under the current Windows user account. The keys are not uploaded to exployt cloud. For sensitive work you can avoid cloud AI providers and choose local models such as Ollama.

The exact categories are listed in our Privacy Policy. In short, exployt cloud processes account, authentication, billing, subscription/license, security, fraud-prevention, support/feedback, and consented website analytics data. We do not collect Big Data from your projects, harvest project files, or build datasets from your source code. Source code, file contents, local databases, agent history, and project details stay on your machine during ordinary desktop-app operation. Marketplace publishing and support submissions are deliberate actions: Marketplace uploads send only the files and metadata you choose for sharing or sale, and support/bug-report flows show you the diagnostic data first and send only what you approve.

No. As of July 8, 2026, exployt does not claim ISO 27001 certification, SOC 2 Type I, SOC 2 Type II, or an equivalent external security certification or audit report. We are preparing the ISMS scope, policies, risk assessment, vendor review, access-review evidence, incident-response evidence, and other audit materials needed for a future ISO 27001 or SOC 2 readiness process.

exployt.ai is operated by exployt Software GmbH, Lange Gasse 30, 8010 Graz, Austria. Commercial register: FN 675480d (Landesgericht für ZRS Graz). VAT ID: ATU83180825. Managing director: Michael Martin Schöggl. Contact: office@exployt.ai.

The primary production web service is hosted on Microsoft Azure in the West Europe region (Netherlands). Your source code, local project files, local project databases, local agent history, and local AI-provider API keys stay on your machine during ordinary desktop-app operation. If you intentionally publish an item to the exployt Marketplace, selected files and metadata for that item must be uploaded to exployt servers so the item can be reviewed, hosted, shared, sold, and delivered to other users. Marketplace items can include integrations, modules, standalone apps, settings sets, workflows, connector packages, or other reusable assets; you choose what to include, and we recommend minimization plus obfuscation where appropriate. For support, feedback, and bug reports, the app or website shows the data to be submitted before sending and submits only what you confirm.

Business and enterprise customers can request a DPA by contacting office@exployt.ai. We are preparing a public DPA package for publication in the coming weeks. Until that is published and signed where required, business customers should treat the signed agreement or separate written agreement as controlling.

For the public website and account service, current service providers include Microsoft Azure, Stripe, Resend, MaxMind offline geolocation data, and optional Google Analytics 4 / Microsoft Clarity website analytics where enabled after consent. This subprocessor list is about exployt cloud services, not every local desktop tool or open-source library. The desktop app itself uses local tools and external services only where you install, configure, connect, or select them, such as Git, local Ollama models, selected AI providers, Marketplace modules, settings sets, or connector packages. Customer-selected AI providers are chosen by you for your local project work and are not exployt cloud subprocessors for that project content.

Passwords are stored as bcrypt hashes, not plaintext. Traffic uses HTTPS/TLS, which is the standard transport security layer for web and API traffic; modern deployments use TLS 1.2 or TLS 1.3 rather than inventing a custom transport cipher. Selected sensitive server-side fields use AES-GCM encryption and lookup hashes. Local Windows secrets such as cached entitlement tokens, credential-vault entries, and build/deploy credentials use Windows DPAPI under the current Windows user account. Selected service secrets and signing material are managed through Azure Key Vault where configured.

Support requests sent to office@exployt.ai or submitted through product support/feedback forms receive an initial response within 24 hours. exployt cloud is needed for account creation, login/account management, billing, payment, subscription/license validation, Marketplace cloud actions, and fresh entitlement checks. Ordinary desktop project work is local-first: the desktop app keeps a signed entitlement snapshot with a maximum 72-hour grace window, so transient server outages do not immediately remove access to subscription-locked features. If an exployt cloud outage occurs, our operational commitment is to restore the account/subscription service before that 72-hour grace window expires. This public commitment does not create service credits unless a separate written enterprise agreement says so.

Your local project files, source code, agent history, local credentials, and local project database remain under your control. The project database is a local SQLite database stored under the project's .exployt folder; when you have not enabled local database encryption, it can be inspected or exported with ordinary SQLite-compatible tools. exployt also provides export/import flows for workflows, settings sets, and many configuration assets such as agent/workflow settings, so you can back them up, share them, or use them outside one installation. The Terms do not restrict you from using your own local project data with other software. For cloud account data and personal data processed by exployt, you can request access, deletion, restriction, or portability under the Privacy Policy and applicable law. Some records must be retained for legal, tax, fraud-prevention, dispute, or security reasons.

exployt cloud does not collect, review, or train on your source code, file contents, or project details during ordinary desktop-app operation. AI work is different from exployt cloud collection: when you explicitly configure and use a cloud AI provider such as OpenAI, Anthropic, OpenRouter, Google Gemini, xAI, MiniMax, Qwen, Claude Code, Codex, Cursor, or Windsurf, the prompts and project context needed for that AI task are sent to that selected provider because cloud AI cannot work without receiving the task context. You control which provider and model is used. exployt supports direct providers, OpenRouter with 350+ models, and local Ollama models; for sensitive projects you can route work to local models so the AI task context stays on your machine.

Separate from AI provider use, you may deliberately publish Marketplace items to share or sell them. Those uploaded files and metadata are transmitted for review, hosting, delivery, and sale. We recommend package minimization, secret removal, and obfuscation for code-containing Marketplace items, and exployt can support that workflow. For support and bug reports, there is no hidden background upload: when diagnostic data may be included, the app shows what will be sent and asks for confirmation. Only the data you approve is submitted.

The desktop app may derive stable device-fingerprint signals from hardware and operating-system attributes. We reduce those signals to hashes/HMAC values and use them only for account security, license integrity, known-device recognition, and trial-abuse prevention, not analytics, marketing, advertising, or cross-service tracking.

Send security reports to office@exployt.ai with enough detail to reproduce or assess the issue. Do not include secrets, passwords, or data that is not necessary for the report. The machine-readable security contact is also published at https://exployt.ai/.well-known/security.txt.

exployt was built by a single person in about five months — started around Christmas 2025 and live by the end of May 2026. The remarkable part: exployt built itself. From the very first minute, earlier versions of the app were used to develop the next ones, so the tool grew by being used on its own creation. It's the app that developed itself.

Essentially none of it. The founder is a self-taught hobby programmer — no formal degree — who wrote virtually no code by hand. The AI did the building, and not just the programming: everything around it too, from company formation to day-to-day management. That's the whole proof of concept. If it works that way for one person with no professional background, it works for you. Coding knowledge is an advantage, but it isn't required.

First and foremost, individuals — the classic one-person shop. People who lost a software or product job to AI, or fear they might, and anyone who wants to seize what may be the single biggest opportunity in modern history.

For the first time, one person can match what used to demand a large funding round and a big team. A good idea, good taste, attention to detail, and real care are now enough to build a product that can outclass a soulless one from a giant company. exployt is for the people who want to use that opening — whether that's a few months on an ambitious project or a few weeks on a small one.

AI can be enormously good or enormously bad for people, and a lot of it comes down to one question: do a handful of large companies build AI monopolies while everyone else just loses their job — or does the opposite happen?

Our philosophy is the opposite. The historic advantage of big companies was funding: more money meant more people on a project. AI collapses that advantage. A single person can now achieve what once took a large organization — and the ordinary individual can win back a share of the value. We're convinced large companies are building tools like exployt internally and have little reason to release them, because that kind of automated, around-the-clock AI workforce is power. So we're building the counterweight: making sure private individuals get the same capabilities, tools, and reach as the big players.

A few things we intend to keep. Your data stays yours: we don't harvest your code or project content, now or later. We keep the focus on empowering individuals against large organizations — that philosophy isn't going to drift. And we keep pushing to make exployt as powerful as we can while keeping it publicly available to everyone.

Concretely on cost: we'll do our best to keep pricing affordable and financeable for individuals over the long term, because the mission is genuine equality of opportunity. Business pricing is a separate matter — it's higher, and we expect it to rise, partly to protect individuals' ability to compete with large companies.

The idea goes back to 2022, when ChatGPT became a phenomenon almost overnight. The assumption was that big companies were already building something like this and would release it any moment. Years passed and it never appeared — and the tools that did show up, like Cursor and Windsurf, didn't meet expectations or exhaust the potential of multi-agent setups. They tended to extend or integrate existing development tools rather than rethink the whole thing from the ground up.

The turning point came when Claude Opus 4.5 landed (around October 2025) and AI suddenly got dramatically better. It became clear that, rather than finishing other software projects by hand or with a human team — and given how hard those projects were to fund — it would be faster to spend a few weeks or months building the right tool: an AI team that solves those problems on its own and then keeps working like a full company. Work began at Christmas 2025.

The name ties to the slogan: exploit the biggest chance in history. The thought behind it is that our current system is unfair to the ordinary person — workers get used, large companies write the rules and set the direction, and many people feel stuck on a treadmill, a bit like the Matrix: rigid routines, little freedom, little real ownership.

AI is the crack in that system. An exploit, in the original sense, isn't a hack and isn't illegal — it's finding and using a weakness that's part of the system itself, exactly what the powerful have always done. We want to turn that around. AI makes things possible that simply weren't before, and the smart move is to use that opening — legally, on your own terms. exploit AI: seize the chances AI creates.

The logo is a dollar sign raised to the power of X — $ˣ — where X is the number of AI agents working for you in parallel. The more capable agents you can run at once and steer well, the more value you create, the better and more valuable your software, and the more you can earn or turn into independence. That's the real exploit: not AI by itself, but managing as much AI as possible in parallel, working together, without constantly needing you. exployt — our app — is the exploit. It's how one person can field hundreds of AI "colleagues" and match a large company. The dollar sign isn't really about money for its own sake — it stands for financial freedom, independence, and a way back to mattering in a system built around capital.