Additional terms for LitoCode
LitoCode is a Windows coding agent that can read and modify files, run commands, and send project context to third-party artificial intelligence providers. It is offered as a staged-access beta: downloading it does not enable use; each account requests access and an administrator decides whether to approve it. These terms supplement the LitoAI Terms of Service, Privacy Policy, and Acceptable Use Policy.
Last updated: August 2, 2026Staged-access beta
Beta access is personal, revocable, non-transferable, and separate from any LitoAI subscription. A download or application does not create a right to access, support, or continuity. Each approval has an expiry date and may include usage limits.
You may request access if you are at least 18 and have legal capacity to accept these terms. When requesting it, you provide your country and intended use and confirm that you have the right to use the projects and data you open. Administration may approve, reject, limit, renew, or revoke access based on capacity, security, and beta suitability.
A Pro plan may receive review-queue priority. It does not purchase LitoCode, bypass evaluation, guarantee approval, or alter the contracted benefits of the LitoAI plan.
Use LitoCode only with projects, data, and services for which you have sufficient rights and authorization, including authorization for processing by the declared providers. Do not use it to operate critical or high-risk systems.
The experimental nature of the beta does not remove any non-waivable rights you have under applicable law.
What LitoCode does on your computer — read this
LitoCode is an autonomous agent, not only a suggestion editor. Within the task you request, it decides which files to read, which changes to propose, and which tools or commands to use.
Its built-in file tools limit writes to the project folder and protect sensitive paths. That boundary does not contain the terminal: commands, scripts, package managers, and external programs run with your Windows user permissions and may read, create, modify, delete, or transmit data outside the project.
The current Preview has no operating-system sandbox. It can access the Internet, install dependencies, start processes and local servers, and open previews. Use it from a non-administrator account with minimal permissions.
What leaves your computer — read this too
To answer, LitoCode sends the turn's working context to our backend and external providers. That context may include the full contents of files it read, command output, repository paths and symbols, instructions, local memory inserted into the turn, and the conversation.
You choose the folder, but you do not approve files one by one: the agent decides what context it needs. Any file the agent can read in the project may be transmitted. Commands may also read other locations, and their output may be placed in context.
Do not use LitoCode with third-party code, data, secrets, or material unless you have sufficient rights and authorization to process it and share it with the declared providers. If a project has confidentiality, security, or retention requirements that are incompatible with that processing, do not open it with LitoCode.
Each request is capped at approximately 700 KB of context.
Which AI providers process your code
Primary provider: MiniMax API, operated by Nanonoble Pte. Ltd. (Singapore). Its current policy states that data is stored in United States data centers, allows inputs and outputs to be processed, and offers no universal zero-retention guarantee.
Fallback: OpenRouter (United States), which routes to third-party model endpoints. On that path LitoCode requires a Zero Data Retention endpoint and denies data collection per request. ZDR means the endpoint processes content to answer but does not retain it afterwards; it does not mean the content stayed on your computer or was anonymous.
Planning and structure explanation: OpenAI API. We send the necessary request, repository map, memory, and instructions with store:false. OpenAI states that API data is not used for training by default, but may retain content in abuse-monitoring logs for up to 30 days unless the account has approved ZDR controls.
Web search: Tavily receives the query. Image generation: MiniMax receives the prompt, settings, and any reference image. Google Cloud/Firebase authenticates the account, runs the backend, and records usage metadata.
Providers, models, and processing locations may change. We will notify you before a material change that expands the use or retention of content. Your consent covers only your own data and is not a substitute for third-party authorization.
What we store and where
Our backend relays LitoCode requests without storing prompt or response content. It retains usage, model, token, and cost metadata per account, the receipt showing acceptance of these documents, and the minimum beta access request and review data. External providers apply their own processing and retention policies.
Your sessions, project index, project memory, snapshots, and activity log are stored locally on your computer. The agent may include parts of that data in a request when relevant; when it does, those parts are transmitted.
AI providers do not receive your email or name in the LitoCode payload. They receive a pseudonymous session identifier and the technical request content; that content may itself identify you if your files, commands, or paths contain personal data.
We do not use your code to train models of our own.
Safeguards built into the product
Built-in file tools reject paths outside the project and protect certain sensitive files. Commands and external programs are not contained by that boundary.
LitoCode takes snapshots of files it manages directly to help with rollback. They are not backups and cannot reliably undo changes made by commands, scripts, external processes, networks, databases, or remote services.
A classifier attempts to detect deletion, publishing, credentials, external paths, and other sensitive commands for approval, and blocks certain patterns of extreme harm. It is a risk-reduction layer, not an infallible security boundary: scripts, nested commands, or external tools may have consequences the classifier does not anticipate.
The local activity log keeps known commands, decisions, and paths. Secret redaction is best effort and may not recognize every token, password, or sensitive value. Do not treat the log as a substitute for proper secrets management.
Turbo mode
Normal mode asks for approval before commands with effects, except recognized read-only commands. Turbo mode automatically runs commands its classifier does not mark as sensitive or prohibited. It is off by default and is enabled per project through a separate warning and confirmation.
Turbo grants broad autonomy over the project and terminal: commands run with your permissions, may use the Internet, and may cause data loss or exposure. The Turbo indicator remains visible and you can disable it at any time.
LitoCode attempts to continue asking for approval for detected sensitive operations even in Turbo, but it cannot guarantee that it will identify every possible effect of a command, script, or tool. Keep backups, use version control, and apply least privilege.
Code generated by LitoCode
As between you and us, you own the code LitoCode produces for you and we assign you any rights we might have in it. You are responsible for how you use it.
AI-generated code may resemble code that exists in public repositories, including code under restrictive or copyleft licences. We do not guarantee that generated code is original, that it is protectable, or that it does not infringe third-party rights. Reviewing it before shipping, publishing or selling it is your responsibility.
We do not guarantee that generated code is correct, secure, performant or fit for any particular purpose. Review it and test it as you would review code written by a person.
Your responsibilities
Keep backups and version control of anything you cannot afford to lose. The snapshots LitoCode takes are a convenience, not a backup system.
Open only projects, data, and services for which you have sufficient rights and authorization. Do not introduce secrets, credentials, or personal data you are not authorized to process or share with the declared providers.
Review what LitoCode changed before accepting it. The diff view and the revert option exist for that.
Read every command before you authorize it and independently verify its real scope. An AI-generated explanation may be incomplete or incorrect.
Do not use LitoCode to build or operate systems where failure could cause death, injury, or serious environmental, financial, legal, or security harm. Before using changes in production, review and test them with appropriate human and engineering controls.
Unsigned installer
The beta installer currently has no Authenticode signature. Windows may show SmartScreen or an unknown-publisher warning. This does not make the software unlawful, but it reduces automatic verification of its source.
Download it only from LitoCode's official HTTPS page and compare the SHA-256 published with the file. Do not install copies received through messaging, unofficial repositories, or forwarded links. Do not disable system protections to install a copy whose source you cannot verify.
No warranty
LitoCode is provided as is and as available. It is an autonomous agent driven by language models: it makes mistakes, misreads instructions, and can produce changes you did not want.
To the maximum extent permitted by applicable law we give no warranty of accuracy, reliability, availability, merchantability, fitness for a particular purpose or non-infringement, in relation to LitoCode, to the code it generates or to the commands it runs.
Nothing in these terms limits rights that cannot be waived under mandatory rules of your country of residence.
Liability
The limitation of liability in the LitoAI Terms of Service applies to LitoCode to the maximum extent permitted by law. In determining the cause and scope of any damage, your instructions, permissions, approvals, lack of backups, use outside the beta scope, or breach of these terms may be taken into account.
These provisions do not exclude liability for failures of the service where the law does not permit exclusion, or for wilful misconduct, gross negligence, fraud, death or personal injury, or non-waivable consumer rights. A warning or confirmation does not create absolute legal immunity for either party.
Access and suspension
Administration reviews applications manually and may limit, suspend, or revoke access for safety, abuse, breach, expiry, operational capacity, or closure of the beta. There is no service-level agreement or promise of availability, support, approval, or continuity.
LitoCode is under active development. Features, models, and limits may change or disappear between versions. A material change to these terms or privacy practices requires new acceptance before you continue using the beta.
Ending the beta does not change or reduce the benefits of any paid LitoAI subscription, because LitoCode is not part of those benefits. Queue priority ends when the Pro plan or the beta ends.
Everything else
The LitoAI Terms of Service, Privacy Policy, and Acceptable Use Policy apply to LitoCode, including their provisions on accounts, content, output, acceptable use, dispute resolution, governing law, and jurisdiction. Pro review priority does not turn LitoCode into a paid benefit or a subscription-included product.
Where these additional terms conflict with the general Terms, these additional terms prevail for LitoCode only.
Before opening a project, the app requests separate acceptance linked to your account, language, app version, and the revisions of these terms and the Privacy Policy. We keep a local receipt and a minimal backend receipt showing which version you accepted and when.
Questions about these terms, privacy, or the access beta: privacy@litochat.com.