A Abe Stack

Trust center

Security & data handling

How we design agentic systems and production web apps for teams that need auditability, approval gates, and infrastructure they control — without black-box SaaS lock-in.

We do not claim third-party certifications we have not earned. This page describes our engineering practices and the platforms we build on.

Human-in-the-loop by default

Outbound agent actions — emails, publishes, roster changes — queue for coordinator approval before execution. Production systems include Firestore audit logs, session traces, and eval checkpoints. See our Coordinator Ops Agent and Inquiry Triage Pipeline case studies.

Client-owned infrastructure

You own the source code, Firebase/GCP project, and deployment pipeline. We hand off GitHub repos, environment documentation, and runbooks — not a rented widget. No per-seat AI SaaS markup on your workflows.

Encryption & access control

Production builds use TLS in transit. Firebase and Google Cloud provide encryption at rest for Firestore, Cloud Storage, and managed secrets. Role-based access, Firebase Auth, and least-privilege service accounts are standard on platform engagements.

Observability & change control

Staging environments before production cutover. Cloud Logging, structured error handling, retries on queue workers, and version-controlled prompts and agent configs. Weekly written status and shared project boards on larger engagements.

AI & LLM data handling

  • LLM providers (Google Gemini, OpenAI, Groq, DeepSeek) process prompts only as required for your workflow — configured per client project, not shared across clients.
  • RAG knowledge bases are scoped to your documents and stored in your Firestore or vector store — not used to train public models.
  • API keys and secrets live in environment variables or GCP Secret Manager — never committed to source control.
  • Demo endpoints on abestack.com use fictional data only and do not call production client systems.

Subprocessors & platforms

Typical production stack components. Client projects may add or remove services based on scope:

Google Cloud / Firebase Vercel Google Gemini OpenAI Groq Calendly Zoho Mail

Website privacy details: Privacy Policy →

Procurement & security questionnaires

For multi-workflow, multi-team, or regulated environments we deliver under fixed-scope SOW with architecture documentation, threat-model lite, and UAT checklists. We can complete vendor security questionnaires and provide architecture diagrams for your IT review.

Discuss an enterprise engagement