FAQ
Frequently asked questions
Can Sigma run entirely inside our own infrastructure?
The Sigma platform deploys on-premise on infrastructure you control, or inside a sovereign cloud region if you prefer not to self-host. Today it still calls out to your chosen model provider and to a managed retrieval index; fully self-contained deployment is on the roadmap. Tell us your requirements.
Does data ever leave our chosen jurisdiction?
The platform, its data store, credentials and audit trail stay where you deploy it. Model calls go to the provider you choose (OpenAI or Anthropic today), and document retrieval currently uses a managed index, so those two follow the provider’s own residency options. In-region and self-hosted options for both are on the roadmap.
How does Sigma prevent an agent from doing something it shouldn’t?
Two mechanisms: access control (an agent’s tools, connectors and data access are explicitly scoped, never open-ended) and human-in-the-loop approvals (a configured set of tool calls require a person to approve before they execute, and a session can be halted mid-flight).
Are credentials ever exposed to the model?
No. Credentials are stored in an encrypted vault and resolved by the platform only when a tool call needs them, and each use is recorded. Raw model keys and people’s passwords never enter the agent’s sandbox.
Can we audit what an agent has done?
Yes. State-changing actions are attributable and timestamped, sessions can be inspected event by event, and administrators can request an explanation of any user’s effective access.
What happens if we want to switch LLM providers later?
Model providers are configured per organization and selected per agent. Switching a provider or model is a configuration change, not a platform migration.
How is Sigma priced?
There is no public price list. A free account is available by invitation (request access), and its limits are shared on invitation. Enterprise, for on-premise or sovereign-cloud deployment, is quoted in a proposal for your organisation. See the Pricing page or contact us.
How does Sigma handle Arabic-language use cases?
Agent behavior in Arabic depends on the language capability of the LLM provider and model you configure; the platform itself places no restriction on the language an agent operates in.