The Sovereign Answer Engine

Defensible answers your CISO will actually approve.

Aeira organizes your organization's knowledge and turns it into defensible answers for the right people and agents. Every answer is identity-bound, provably erasable, and audited. Deploy on AWS, Azure, GCP, on-prem, or fully air-gapped.

Open full slides ↗
Design Aeira Deployment Pricing & Tiers Architecture Slides
  • Identity-bound: the agent never sees what the user can't
  • Provably erasable: for GDPR / HIPAA obligations
  • Audited: it carries the audit trail your compliance team can actually point to
  • Especially relevant for: defense, finance, healthcare, and federal; the same ACL-aware data plane works for any enterprise that wants its knowledge inside the firewall
Architecture details available under NDA
Aeira evaluates permissions before retrieval. Three colleagues run searches against the same corpus: Finance asks for a supplier invoice and is served it, Engineering asks for an architecture diagram for Project Northwind and is served it, and Sales asks for the same diagram and gets No access, because the document was never a candidate for that identity rather than filtered out afterwards. The poster's figures are from a single-machine run on a synthetic 5.4 million document corpus loaded through the bulk API: 64 concurrent clients, 82 searches per second at a 591 ms median, and zero authorization violations across 14,013 returned documents checked.
The same search, three identities, three different result sets. The Sales result is not a hidden row: the document never entered the candidate set, because permission is evaluated before retrieval rather than after it. Measured on one on-prem machine (32 cores, 91 GB RAM) against a synthetic 5.4 million document corpus loaded through the bulk API, not through connectors, with ACLs enforced and content encrypted at rest. Keyword lane, 64 concurrent clients: 82 searches per second at a 591 ms median, and zero authorization violations across 14,013 returned documents checked. A single-node measurement, not a cluster figure and not a projection. An earlier, higher throughput figure had no surviving record and was withdrawn.

Capabilities

Why Aeira

Core capabilities that drive measurable enterprise impact.

Identity-bound access control

Every query is automatically filtered to what the calling user's identity is entitled to see. That is what makes it defensible: access-correct, cited, and reproducible.

  • Derived from your existing IdP: region, department, sensitivity clearance, and role
  • Your AI agents and your auditors get the same answer, filtered to the same identity

Provable erasure for GDPR / HIPAA

Per-tenant crypto-shred makes a whole organization's encrypted content unreadable with a verifiable receipt. One person's erasure request is a per-person delete, with its own receipt.

  • Auditor-acceptable
  • Independently verifiable

Air-gap capable

All indexing, search, embedding, and AI retrieval run inside your security perimeter.

  • Deploy on: AWS, Azure, GCP, on-prem, or fully air-gapped
  • No telemetry; close egress and use a model you host to run with no outbound calls

Stable API. Standalone or with Bolt.

Same contract from Catalog (free with Bolt) through Federated (air-gapped enterprise).

  • Scale tiers without integration rewrites
  • Customers with their own AI layer can buy Aeira alone; everyone else gets it bundled

FAQ

Questions?

What makes an answer "defensible", and does Aeira generate it?

A defensible answer is one you can stand behind in an audit: filtered to the caller's identity (they never see what their ACL forbids), cited to its source, and reproducible. Aeira produces that access-correct, cited evidence and its provenance. When you want it synthesized into prose, that generation runs on your own BYO model inside your own boundary, with PII masked at the model boundary; Aeira itself makes no outbound model calls.

Is Aeira Catalog included in all Bolt plans?

Yes. Catalog is bundled free in every Bolt plan as the baseline data plane. Dynamic, Enhanced, and Federated are paid Aeira tiers for larger scale and stronger governance.

What does air-gapped deployment mean in practice?

Federated-tier deployments run entirely inside your security perimeter, typically a VPC, on-prem cluster, or physically air-gapped network. No outbound calls, no telemetry, no third-party model API calls. License validation happens via an offline-signed token, refreshed on a customer-controlled schedule. Catalog and Dynamic also run fully on-prem; Federated adds dedicated cluster isolation and custom topology.

How does Aeira help with GDPR / HIPAA erasure obligations?

Two mechanisms. Crypto-shred is per organization: destroying the tenant's key makes its encrypted content mathematically unreadable, with a verifiable receipt, which fits offboarding or disposing of a whole tenant. One person's Right to be Forgotten request is handled by deleting that person's records and the documents about them, not by a key. Some data, such as sealed audit records, is retained by design. We can walk through the specifics in an architecture review under NDA.

Can we bring our own data, embedding model, and infrastructure?

Yes to all three. Customer pays for their own compute, storage, and infrastructure. Aeira ships with sensible defaults but supports BYO embedding models for customers with specialized domain models. The access-control semantics stay consistent regardless of which embedder is used.

What changes when we upgrade tiers?

Scale and governance depth, not the API. Catalog is keyword-only with hard record limits. Dynamic adds full hybrid search up to 10K records. Enhanced supports up to 100K records with high-throughput ingestion. Federated adds air-gap, multi-region, and dedicated cluster isolation. Applications written against Catalog scale up to Federated without rewrites.

Pricing: what does each tier cost?

Catalog: free with every Bolt plan. Dynamic: $999/month (up to 10K records). Enhanced: $4,999/month (up to 100K records, high-throughput ingestion). Federated: custom annual contracts from $500K/year (defense / federal / regulated tier). All paid tiers include software updates, patches, and support; customer pays for their own compute, storage, and infrastructure.

Can we see the architecture and security details?

Yes, under NDA. Detailed architecture briefs, encryption specifics, key management semantics, and audit-trail formats are shared during pilot evaluation. Schedule an Aeira deployment design call and we'll cover what's relevant to your environment.