Atlas questions, answered honestly.

Yes. Atlas supports CSV, Excel, and JSON import for applications, technologies, capabilities, and decisions. There is a five-step import wizard with column mapping, validation, and preview before you commit.

We use Anthropic Claude — Haiku for classification and simple extraction, Sonnet for most analyses, Opus for strategy parsing and capability generation. The model is visible in the Langfuse trace on all AI outputs.

Data is stored in Frankfurt (EU) via Neon PostgreSQL. Anthropic zero-retention policy is active: prompts and outputs are not stored by Anthropic after the API response.

Yes. Native connectors for Jira, Azure DevOps, ServiceNow, GitHub, Confluence, Notion, Slack and more are live, alongside REST API, webhooks and MCP. The canonical status matrix lives at /integrations.

Yes. CSV, JSON, and PDF export is available for all entity types. Data is deleted completely within 30 days of cancellation — export is always possible before then.

Most teams have a working portfolio within one working day: import existing data, AI enrichment, first capability map. We offer a two-hour onboarding session for new customers.

We actively work to GDPR principles: DPA is public, sub-processor list is published and updated on changes, data subject rights are supported (deletion, export). We have no external certifications yet — we state this honestly.

All data is stored in Frankfurt, EU (Neon PostgreSQL, Vercel fra1 region). No data leaves the EU without explicit consent.

Yes. You can export and delete your workspace at any time. On cancellation, all data is deleted completely within 30 days. Individual data objects can be deleted via UI or API.

Every AI feature runs behind a server-side allowlist guard, with a monthly cost ceiling per workspace, a full trace on every call, EU AI Act Annex III risk classification on AI use cases, and a human review flag on high-risk classifications.

Not yet — we are pre-certification. SOC 2 Type I preparation is in progress (target Q3 2026); ISO 27001 is on the roadmap (Q3 2027). We publish our security evidence library and policy stubs at /trust — including what we have and what we don't.

Atlas is free during early access, at least until October 2026. It becomes a paid product later. We do not publish price figures on spekir.com yet.

Atlas is free during early access. There is no self-service trial; you ask for access on the waitlist, and we invite in waves so every new user gets a proper start. Sign up at /early-access.

Write to hello@spekir.com — we respond within one working day and find a time that works. Demos are typically 45 minutes with a live walkthrough of your use case.

Spekir was founded in Aarhus, Denmark. We are a small, deliberately lean team — the founder still writes code. Write to hello@spekir.com for anything.