Security & trust

You are handing agents access to real accounts

That deserves a straight answer about where your data goes, what an agent can touch, and what we have not built yet.

Where your data lives

Your agents, chats, memories and documents are stored in a managed Postgres database with per-row access control, hosted on Azure.

Every table that holds your data is protected by row-level security, and every query is scoped to your user and project.
Files your agents produce or read are stored in Azure Blob Storage in dedicated containers, not alongside other tenants’ application data.
Backups and the database itself are encrypted at rest by the platform; all traffic to the app is over TLS.

The tokens that let an agent act on your behalf are encrypted before they are stored, with a key the database never sees.

Integration secrets are encrypted with AES-256-GCM, an authenticated cipher, so tampering is detectable rather than silent.
Encryption keys and all platform secrets live in Azure Key Vault and are delivered at runtime. Nothing is baked into the application image.
We never see or store your account passwords — authentication is handled by Supabase Auth.
Rotating the encryption key deliberately invalidates stored secrets: affected integrations must be reconnected rather than silently re-read.

Connecting a tool and authorising an agent to use it are two separate decisions. Connecting Slack does not give every agent Slack.

Each agent holds explicit per-integration grants. An agent with no grant for a tool cannot call it, regardless of what it is asked to do.
Revoking a connection drops every grant that depended on it, immediately and everywhere.
Outward-facing actions — publishing a post, sending a message, changing ad spend — are gated behind an approval step you see before anything leaves your account.
Agents run against a token budget you control, so a misbehaving loop stops rather than running up a bill.

Running an agent means sending its context to a language model. We run one, and we run it ourselves.

Runs use the platform model we provide, hosted on Azure AI Foundry within our own infrastructure. Bringing your own model or API key is not supported.
Because there is a single provider, there is no question of your content reaching a third-party endpoint you did not expect.
We do not train models on your data.

Enough to run the service and show you what your agents did, and no more.

Agent runs are traced so you can see which tools were called and what each step cost.
Operational logs capture errors and metrics for reliability work. Secrets are scrubbed before anything is written.
Token usage is recorded per run for billing and for the budget gauge in your dashboard.

Your data stays until you remove it.

Deleting an agent removes its configuration, its memories and its grants.
Disconnecting an integration deletes the stored tokens for it.
To delete your account and everything attached to it, email contact@selfagentic.in and we will action it.

Certifications get claimed loosely in this category. Here is where we actually stand.

We are not SOC 2 or ISO 27001 certified. We are a closed-beta product and will not claim otherwise.
We have not completed a third-party penetration test.
SSO/SAML and role-based access are built for enterprise plans; talk to us about your requirements before relying on them.

Security reports are read the day they arrive. Review our Privacy Policy and Terms for the legal detail.