FAQ

Where does my data live?

Production runs entirely in AWS eu-west-1 (Ireland). Indexed chunks live in OpenSearch Serverless, connector secrets in AWS Secrets Manager, audit log + metadata in RDS Postgres. We never replicate customer data outside the EU. Outbound calls to Stripe (US) and WorkOS (US) carry only billing / authentication metadata, never customer documents.

Do you train on my content?

No. Embeddings are generated via AWS Bedrock Titan with logging=disabled; nothing leaves the AWS account. We don't run any model training or fine-tuning on customer content, ever.

Which models do you call?

Gigamcp itself only calls embedding models (Titan Embed v2). The AI assistant on the other end of the MCP socket is whatever you choose — Claude, GPT, local models. Gigamcp doesn't proxy LLM completions.

Can I bring my own OpenSearch / vector DB?

Not on the multi-tenant SaaS. Enterprise customers can opt into silo-mode (a dedicated AWS account in your region of choice with your own AOSS collection) — talk to sales.

Do you support SSO?

Email + magic-link login is built-in. SAML / OIDC SSO and SCIM provisioning are an Enterprise feature, delivered via WorkOS. Setup takes ~30 minutes once we receive your IdP metadata.

How do you handle deletion?

Workspace owners can delete the workspace from Settings → Danger zone. We mark it for deletion immediately (the API stops responding within seconds), purge OpenSearch and Secrets Manager within 24 h, and hard-delete Postgres rows after a 30-day grace period (so you can recover from an accidental click).

What if I hit my plan limit mid-conversation?

The MCP request returns 429 with a clear message; the AI client usually surfaces it to the user as "rate limit reached, try again in N seconds". Bumping to the next plan takes effect immediately.

Where do I report a security issue?

security@gigamcp.io. PGP key on the same page. We aim for an initial reply within 24 h. Public bounty program is on the post-launch roadmap.

Status page?

status.gigamcp.io. Powered by an external provider so it stays up when we're down.