Continuum Memory API · Legal

Privacy Policy

How Syntarus handles data for the Continuum Memory API and developer console.

Last updated: 10 October 2026

1. Scope and contact

Syntarus operates Continuum, a memory API and developer console for applications and AI agents. This policy covers our website, console and hosted Memory API. Dedicated or customer-hosted deployments also follow the signed agreement. Privacy requests: [email protected]. Syntarus is based in India; postal correspondence details are available on request.

2. Information processed

  • Account data: name, email, profile image and account identifier supplied during sign-in, plus session records.
  • Project data: configuration, scoped credential records, user/agent identifiers and usage/plan records.
  • Customer content: messages, documents, events and metadata your application submits; retained source evidence; derived memories, embeddings, relationships and state cards where enabled.
  • Operational data: request identifiers, timestamps, status, usage counters, error/security records, webhook delivery records and ingestion receipts.
  • Support communications: information you send us for integration, billing or support.

Send only the content needed for your workflow. Do not submit passwords, private keys or information your application is not authorised to process.

3. Customer and end-user responsibilities

A customer decides which end-user information its application sends, the purpose of processing and the project/user/agent scope. The customer must provide appropriate notices, establish the required legal basis and obtain permissions where necessary. End users should contact that application first; we support authorised customer requests affecting the Memory API.

4. Processing modes and AI providers

Hybrid ingestion retains source evidence and uses model processing to extract memories and relationships. Direct and document ingestion can store/index submitted content without model-based fact extraction. Embeddings and retrieval are separate from extraction. Consult the API reference for the mode you choose.

Hosted model extraction uses Sarvam AI. The platform also supports configured Google Gemini/Vertex AI processing or fallback; the applicable configuration determines which provider receives content. A model-enabled write may send submitted content and relevant existing context to the configured provider. Your application's answer model is a separate service you choose.

We do not sell customer content to advertisers or use it to train our own generative models. Third-party provider retention, processing location and training restrictions depend on provider terms and the applicable account agreement. We do not give a blanket no-training or India-only processing guarantee for every provider configuration. Confirm processing terms with us before submitting sensitive or regulated information. See Sarvam's privacy policy and Google Cloud's privacy notice.

5. Purposes, access and sharing

We process information to authenticate accounts, operate storage/retrieval, enforce project access and usage limits, deliver ingestion/webhook status, provide requested support, prevent abuse and meet legal obligations. Operational measurements help diagnose failures and service performance.

Service providers process information for hosting, network delivery, sign-in and configured AI processing. Hosted infrastructure uses Oracle Cloud; Google provides console sign-in. Authorised support or operations access is for service, security and legal purposes. Legally required disclosures may also occur. Contact us for processing details needed for your deployment agreement.

6. Sign-in, cookies and website assets

Google sign-in provides identity information for the developer account. It does not grant Continuum access to your Gmail inbox or Calendar. Session cookies support authentication; local browser preferences support the console. Public pages load Google Fonts, and Google sign-in loads Google's identity components; those providers receive the network information needed to serve requests.

Signing out or clearing cookies does not delete project data. We do not offer advertising targeting based on customer memory content.

7. Storage and retention boundaries

Account, project, source, memory, graph and operational records are stored in platform databases. Provider processing and contracted deployments may have different locations and retention terms; request confirmation of residency before relying on it.

Project retention settings apply to the memory/evidence lifecycle. A memory TTL is not a promise to erase every ingestion-job payload, receipt, account record, audit record or backup. Deleting one memory is not the same as purging all information for an end user or closing an account.

Operational records and backups can remain after active-store deletion for recovery, security, accounting or legal needs. We do not promise a universal backup-erasure deadline here. A specific schedule must be agreed for workloads requiring it.

8. Access, export and deletion

Project-authorised users can inspect/export supported memory records and use memory or durable end-user deletion controls. Deletion jobs are asynchronous: retain the receipt and check completion and affected-store results rather than treating acceptance as completion. Export does not necessarily include every internal operational record.

For project/account closure, broader source-payload removal, privacy rights, correction or a deletion confirmation beyond the API receipt, contact [email protected]. We verify authority and handle requests under applicable law and any signed agreement. Revoking an API key stops its access; it does not delete data.

9. Security and suitability

Controls include HTTPS, project-scoped access, revocable keys, signed webhooks and monitored dependencies. Encryption of sensitive secrets does not imply customer-controlled encryption of every record. See the trust center for implemented controls and certification, residency and SLA boundaries. No system is perfectly secure.

The developer service is not directed to children. Customers serving minors or processing sensitive/regulated information must establish suitable safeguards and confirm service suitability before ingestion. We notify affected parties of incidents as required by law and contract.

10. Changes and questions

We identify the policy revision date and provide notice of material processing changes through reasonable available channels or as required by law or contract. Questions and rights requests: [email protected]. Read the Memory API Terms of Service alongside this policy.