Service Terms

1. How these Service Terms apply

These Service Terms supplement the Terms of Use. They apply according to the capabilities you use, regardless of which Rainspeed Labs product or interface provides them. They do not create a separate legal provider for each product.

If these Service Terms conflict with the general Terms of Use for a specific feature, these Service Terms control for that feature. A signed agreement, order form or checkout disclosure controls where it expressly states different terms.

2. AI and generated Content

AI features may route requests to different models or providers and may combine model responses with search results, files, tools or system instructions. Model availability, context limits, modalities, latency and Output may change as providers and configurations change.

Review generated Content before use. Code may contain vulnerabilities or incompatible dependencies; research may contain incorrect citations; generated media may resemble existing works or people; and generated files may contain formatting or calculation errors.

You are responsible for permissions for reference images, voices, likenesses, music, files, datasets and other Input. Do not represent generated Content as human-created where that representation would be deceptive or unlawful.

3. Search and retrieved information

Search-enabled features may send a query to Exa, Tavily or another configured search provider, and may use ScrapingFish or scrape.do to retrieve selected public pages. Search results, citations and retrieved pages can be incomplete, outdated, unavailable or incorrectly interpreted.

You must open and verify important sources before relying on a response. Rainspeed Labs does not control third-party websites, their content or their continued availability.

4. Agents, tools and execution

AtlasFlux agent features may plan steps, create or edit files, call tools, generate artifacts, use a temporary E2B sandbox or call connected services. ONeNas Code operates locally and can access files, terminal and network resources allowed by the device; it is not sandboxed by default.

An agent plan, permission prompt or completion message is not proof that an action was safe or successful. Review commands and diffs, use least privilege, maintain backups and inspect results. You are responsible for approvals and for effects on your files, repositories, accounts, infrastructure and third-party services.

Context7 MCP and other MCP or connected tools may receive the query and tool parameters needed for the requested operation. Treat every external tool as a separate trust boundary.

5. API and developer features

Developer features may include response and chat-completion interfaces, model listings, API keys, usage records, logs, playground tools, web search and billing. A listed or documented feature may be unavailable, not implemented or restricted for an account or deployment.

API keys identify and charge the account that owns them. Store keys securely, rotate compromised keys, enforce authorisation for End Users and prevent use that exceeds your intended scope or budget.

Requests may be subject to scopes, rate limits, provider limits, moderation, balance checks and idempotency rules. We may return errors, reject requests or change a model route to protect reliability, safety or billing integrity.

6. Local and downloadable software

Local software may work directly with project folders, files, terminal tools and local sessions. Those resources remain under the operating-system permissions and workspace access you provide. A connected model operation may transmit the selected prompt and context needed for a response.

Desktop, CLI and TUI authentication state is stored on the local device. The desktop attempts to use operating-system secure storage; the CLI currently writes token JSON under `.onenas-code`. Keep the device, local configuration and credentials secure.

OpenCode-derived and other open-source components remain governed by their applicable licences and notices.

7. Media generation and processing

Media features may process images, video, audio, music, voices or 3D-related files through configured providers. Processing time, supported formats, file limits, duration, resolution, watermarks and cost depend on the selected feature and provider response.

Some generated media and artifacts may be served through public URLs. Pinned, public, community-shared and unpinned items can follow different storage and cleanup rules. You must have permission to use uploaded media and any person's image, likeness or voice.

8. Storage, history, exports and downloads

Generated results and histories may be stored for a limited period. Certain unpinned generated media and completed or failed media jobs have a 30-day cleanup target. This is not a promise that every file or database record is deleted exactly at that time.

Product interfaces may provide pin, save, publish, export or download controls, but their scope differs by feature. A current export link or key may be intended to expire after 48 hours, and an export may not contain every record or stored object across the ecosystem. Download important files promptly and do not treat generation history as permanent archival storage.

9. Support features and attachments

Support features include Help Center articles, search, known issues, tickets, replies, notifications and staff workflows. Help Center information is general guidance and may not reflect every account configuration.

Where enabled, Support attachments use a private Cloudflare R2 bucket, signed upload and download URLs, server-side ownership checks and file verification. Current limits are 10 MB per attachment and up to five attachments per ticket. Supported formats are limited to PNG, JPEG, WebP, PDF and selected text formats. File validation does not guarantee detection of every malicious file; do not upload secrets or files you are not authorised to share.

Only abandoned unverified staged uploads have a defined 24-hour cleanup eligibility period. Tickets, replies, email records, Help Center activity, audit events and security records may require a rights request or separate retention process.

10. Billing and provider selection

AtlasFlux AI may use Credits and subscriptions, while AtlasFlux API uses a prepaid MYR wallet and may use automatic reload. ONeNas managed runs may use AtlasFlux Credits. Direct ONeNas providers are selected and paid for by the user under the provider's own terms.

A provider selected through a routing service may be operated by another model developer. The selected model, feature and route determine which Content is transmitted and which provider limits apply.

11. Beta, preview and experimental features

A feature identified as beta, preview, experimental or early access may be incomplete, changed without notice, unavailable at times or removed. Do not use such a feature for production, safety-critical or regulated work unless a written agreement expressly permits it.

Data-handling or retention details that differ for a preview will be disclosed in the feature or its documentation where applicable.

12. Third-party terms

A selected model, integration, MCP server, open-source component, direct provider or third-party service may impose its own licence, acceptable-use requirements, costs and privacy practices. You are responsible for complying with terms presented for a third-party account or component you choose to use.