1. Data We Collect
We collect only the categories of information necessary to deliver personalized executive assistance, grounded memory recall, and authorized tool integrations:
Account & Identity Credentials
Your name, email address, and username. Authentication records: WebAuthn passkey public keys and credential IDs (private keys remain exclusively on your physical authenticator); cryptographically hashed passwords with per-user unique salt; and encrypted Multi-Factor Authentication (TOTP) secrets.
Conversational Context & Memory Records
Chat transcripts, user prompts, assistant replies, task goals, and episodic memories. When you explicitly confirm or extract personal preferences, facts, or career milestones, these are retained in your scoped memory store to power future contextual recall.
Connected Service Context
Metadata and authorization tokens from third-party integrations you explicitly connect (such as Google Calendar event schedules, Gmail metadata when syncing correspondence, or Slack workspace identifiers). Tokens are stored encrypted at rest.
Operational Telemetry & Audit Logs
System logs recording sign-in attempts, rate limits, performance metrics, and tool execution status. Audit logs capture authorization decisions and tool names, strictly excluding sensitive user prompt payloads and personal memories.
2. How We Use Your Data & Tenant Isolation
We use your data solely to provide, operate, and maintain Agent Pandu. This includes:
- Retrieving relevant personal context and episodic memories to ground AI agent conversations.
- Executing scheduled background tasks, reminders, and career workflow automations.
- Enforcing per-account rate limits, daily spend controls, and abuse prevention guards.
Multi-Tenant Data Isolation
Every memory document, vector embedding, chat turn, and goal is deterministically scoped by your unique account tenant identifier. Backend authorization middlewares strictly reject cross-tenant retrieval requests. Your data is logically isolated and completely inaccessible to other accounts.
3. AI Model Providers & Zero-Training Guarantee
To generate intelligent analysis and conversation, Agent Pandu communicates with enterprise foundation model providers (such as Google Cloud Vertex AI, Anthropic, and OpenAI) via secure API endpoints.
Zero-Training Guarantee
We maintain commercial and enterprise agreements with our model providers specifying that customer prompts, context memories, and system inputs are never used to train, retrain, or improve foundation models. Your thoughts, work history, and personal context remain exclusively yours.
Contextual retrieval compiles only the minimal useful context necessary for each turn before dispatching the prompt to the inference endpoint.
4. Security & Encryption Controls
We employ defense-in-depth engineering practices to protect your data across every lifecycle stage:
Encryption in Transit
All communications between your browser and our services are protected using Transport Layer Security (TLS 1.3 / TLS 1.2) with strict HTTP Strict Transport Security (HSTS).
Encryption at Rest
Database records and episodic memory storage are encrypted at rest using Google Cloud Key Management Service (Cloud KMS) envelope encryption and Data Encryption Keys (DEKs).
Hardware-Backed Authentication
We support FIDO2 / WebAuthn passkeys (Touch ID, Face ID, YubiKey) so your private authentication keys never leave your client device.
Multi-Factor Authentication (MFA)
Time-based One-Time Password (TOTP) two-factor authentication provides an additional layer of security during account sign-in and sensitive actions.
5. Third-Party Processors & Infrastructure
We partner with vetted infrastructure providers who adhere to rigorous security and compliance standards:
| Provider | Purpose | Location |
|---|---|---|
| Google Cloud Platform (GCP) | Cloud Run hosting, Firestore database, KMS encryption | United States |
| Resend | Transactional emails (account verification, password resets) | United States |
| Model Providers (Google Vertex, Anthropic, OpenAI) | LLM reasoning and inference APIs (zero-training enterprise) | United States |
6. Data Retention, Export & Account Deletion
You have complete sovereignty over your cognitive data. We provide self-service controls to manage, export, or permanently delete your account:
Data Export
You can export your episodic memories, structured facts, and conversation transcripts in standard machine-readable JSON format from Settings.
Immediate Self-Service Account Deletion
You can permanently delete your account at any time in Settings. Account deletion instantly invalidates all active session tokens, removes authentication credentials, and marks user memory and chat records for immediate deletion.
Hard Purge & Backup Expiry
In accordance with operational disaster recovery standards, encrypted database backups are retained on a strict rolling schedule and automatically expire. If you require expedited full-purging covering backup snapshots, email support@agentpandu.com.
8. Regulatory Rights (GDPR/CCPA) & Contact Information
Depending on your location, you may have specific statutory rights under data protection laws such as the General Data Protection Regulation (GDPR) or California Consumer Privacy Act (CCPA), including:
- The right to access and receive a copy of your personal data.
- The right to rectify inaccurate or incomplete records.
- The right to request the erasure of your personal data (“right to be forgotten”).
- The right to data portability.
For privacy inquiries, Data Protection Officer (DPO) requests, or to exercise your rights, please reach out:
Agent Pandu Privacy & Support
Email: support@agentpandu.com