Security, identity, and access
Audit, authorization, isolation, encryption, disclosure, SSO, roles, and what happens when people join and leave.
- What does the audit story look like?Every run records the task, model and tool calls, sources, actions, approvals, and result in an append-only trail that includes every grant and denial.
- How do we get security documentation for a review?Use the request form in the Trust Center or email security@context.ai. Public documents link directly; NDA documents are shared once it is in place.
- Which certifications do you hold?Context does not claim certifications today. A third-party penetration-test report, a CAIQ, and the written security policies are available for review under NDA.
- Is the platform penetration-tested?Yes, by a third party as part of the vulnerability-management program committed to in the DPA. The latest report is available on request under NDA.
- How is data encrypted?TLS in transit and industry-standard encryption at rest, as one item in the written security program in the DPA's Annex 2.
- How are tenants isolated?Your organization is a predicate bound into every query, not a filter applied afterwards, so another tenant's row is never a candidate result. It is identical in all four deployment models.
- How is each agent action authorized?Per action, not per session: deny, then allow, then refuse, with the specific resource bound into the decision. Read, write, and human approval are configured separately.
- How are applets kept safe?An applet renders in an isolated frame with no network access and reaches the workspace only through a narrow bridge that needs a real user click. Backend functions run isolated too.
- Can agents reach the internet from a sandbox?Only as your deployment allows: open, a domain allow-list, or workspace-only. The posture is stated to the agent, and connectors go through the broker either way.
- How do I report a vulnerability?Email security@context.ai with the affected surface, reproduction steps, and the impact you observed. Remediation and disclosure timing are coordinated in that thread.
- How quickly do you notify a data breach?Within 48 hours once a suspected breach is judged likely, within 72 hours once confirmed, and a written summary within ten business days of resolution, per the DPA.
- Do you support single sign-on?Yes. SAML and OIDC are configured in the deployment's setup console with a test sign-in before activation; password login can stay on or be turned off.
- Which identity providers work with Context?Okta and Microsoft Entra ID are the named integrations; any standard SAML or OIDC provider configures the same way.
- What roles exist?Organization roles are owner, admin, and member (agents hold one too). Workspace roles are owner, admin, member, and viewer, with custom roles available.
- How do we invite people?Owners and admins invite by email and can pre-assign workspaces and roles. Deployments can require an invitation to join at all.
- What happens when someone leaves?Removing their membership ends their sessions, revokes their keys, and stops any schedules or triggers that ran on their authority.
- Can an admin see my personal workspace?No. A personal workspace never gains another human member, and an administrator role confers nothing inside it.
- How do agents get their identity?Each agent is its own principal with an organization role. When a run starts it is bound to the agent, the requester, and the policy, and its permissions derive from the people it acts for.
The evidence behind these answers is in the Trust Center: read the deep dive →
Still need an answer?
Tell us what you were looking for and we reply within one business day.
Contact the team →Running a security review?
Documents, controls, and the request form live in the Trust Center.
Open the Trust Center →