From managed cloud to fully air-gapped hardware — pick the deployment model that matches your security posture.
Get started in minutes. Shared cloud infrastructure.
For teams evaluating TARA or working with non-sensitive documents. Hosted in a shared cloud environment — fastest path to a working deployment.
Managed cloud environment. Shared infrastructure. Encrypted at rest and in transit.
Best For
Rapid prototyping, small teams, and non-sensitive document analysis.
Deploys into the cloud you already run.
Billed annually. Unlimited users. Includes 100 GB corpus, 10 team buckets, 500K queries/month.
Self-hosted deployment into your existing Azure, AWS, or GCP environment. Full control over data residency, LLM provider, and scaling. Priced by infrastructure capacity, not per seat — add as many users as your team needs.
Deploys into your own Azure, AWS, or GCP. Your documents and embeddings never leave your perimeter.
Best For
IT teams with existing cloud infrastructure who require control over data residency and LLM provider selection.
| Dimension | Team | Business | Scale |
|---|---|---|---|
| Price | $2,500/mo | $6,000/mo | From $15,000/mo |
| Corpus | 100 GB | 500 GB | Unlimited |
| Buckets | 10 | 50 | Unlimited |
| Queries/mo | 500K | 2M | Unlimited |
| Support | Standard | Priority (4hr SLA) | Priority + dedicated lead |
Approaching your tier limit? We'll reach out 30 days before you hit corpus, bucket, or query caps so you can decide whether to upgrade. No surprise overage charges.
Air-gapped. Hardware included. Zero external dependencies.
Includes hardware, first-year license, and deployment support.
Pre-configured hardware for fully disconnected deployments. Verified no outbound network traffic, no license check-in, no telemetry. Runs indefinitely in air-gapped environments.
True air-gap. No outbound traffic required at any point after deployment. No license check-in.
Best For
Field operations, classified environments, and organizations with strict no-cloud policies.
Every capability, side by side.
| Feature | Cloud | Enterprise | Sovereign Node |
|---|---|---|---|
| Capacity & Scale | |||
| Users | Up to 3 | Unlimited | Unlimited |
| Document corpus | 2.5 GB | 100 GB – Unlimited | Hardware-dependent |
| Team buckets | 1 | 10 – Unlimited | Unlimited |
| Monthly queries | 2,500 credits | 500K – Unlimited | Unlimited |
| Deployment | |||
| Hosting | Shared cloud | Your cloud (BYOC) | Pre-configured hardware |
| Air-gapped deployment | — | — | |
| Infrastructure as code | — | ||
| Kubernetes-ready Helm charts | — | ||
| Data & Sovereignty | |||
| Data never leaves your perimeter | — | ||
| Bring your own LLM provider | — | ||
| Local LLM inference | — | Optional | |
| Encrypted storage at rest | |||
| Identity & Access | |||
| SSO / SAML / OIDC | — | ||
| Team buckets with RBAC | — | ||
| Upload audit logs | — | ||
| Core product | |||
| Source-cited answers | |||
| Persistent agent memory | |||
| Streaming chat responses | |||
| Tri-agent architecture | |||
| Support | |||
| Email support | |||
| Assigned deployment team | — | ||
| 4-hour response SLA | — | Business tier + | Included |
| Quarterly architecture review | — | Business tier + | Included |
Answers to what procurement and security teams ask first.
Cloud tier: your documents are processed in our shared cloud environment and encrypted at rest. Enterprise and Sovereign Node: nothing. Documents, embeddings, and queries stay inside your infrastructure. Embeddings are computed locally. If you configure a hosted LLM provider on the Enterprise tier, query text is sent to that provider — but this is your choice, not a requirement, and local LLMs (Ollama, LM Studio) are fully supported.
TARA's cost to run scales with how much content you index and how many queries you run — not how many people access it. Pricing by infrastructure capacity means you can roll the product out to 50 or 5,000 internal users without the bill changing. It also avoids the seat-utilization and true-up clauses that per-seat enterprise software drags along. If you need more corpus, more buckets, or more query throughput, you move up a tier. Your headcount is your business.
Yes on Enterprise and Sovereign Node. TARA uses LiteLLM under the hood, which supports OpenAI, Anthropic, Groq, Azure OpenAI, Ollama, LM Studio, and any OpenAI-compatible endpoint. You supply the API key or point to a local model. On Sovereign Node, local LLM inference runs on the included GPU hardware — no external provider needed.
The Sovereign Node license is an RSA-4096 signed file verified locally at startup. There is no network call-home, no telemetry, and no entitlement check. The deployment continues to operate indefinitely once the license is installed. License renewal is handled by issuing a new signed file — also transferable offline.
Cloud tier: you can export your documents and memory at any time; data is deleted 30 days after cancellation. Enterprise and Sovereign Node: your data is already in your infrastructure. Canceling means the license no longer renews; the deployment itself remains operational until the license term ends. You retain all documents, embeddings, and exported conversation data indefinitely.
Book a 30-minute scoping call. We'll walk through your compliance requirements, infrastructure, and team size — and tell you honestly which tier (if any) is the right fit.
Book a call