Feith is FedRAMP Moderate Authorized · High In Process Read the release
Feith
Architecture & deployment

Built to run inside your boundary.

Two views of one platform: the capability stack and the deployment topology. Identity, records, AI inference, and audit all stay inside your security boundary. Written for the IT, security, and architecture teams evaluating Feith on the merits.

2 viewsThe capability stack + the deployment topology
3 deployment optionsOn-prem · agency cloud · FedRAMP cloud
8 operating environmentsNIPRNet to AWS Top Secret C2S
100% single-tenantNo shared compute, database, or storage
The capability stack

Five layers. One platform.

Read top to bottom, the five layers trace a record’s path: captured at the top, held in the vault, enriched by AI, driven through workflow, and reached through the access layer. Identity, audit, and compliance aren’t a sixth layer; they wrap all five.

Identity · audit · compliance · wraps every layer
L1
Capture layer Connectors, Records API, ETL, federation, normalization
L2
Records vault Single-tenant store, lifecycle services, holds, retention
L3
AI orchestration Prompt Runner, model-agnostic inference, scheduled execution
L4
Workflow engine Low-code automation, routing, escalation, SLA enforcement
L5
Access layer Applications, APIs, MCP, portals, dashboards
Layer by layer

What each layer does.

Capture layer

Connectors · Records API · ETL

Records arrive from the systems agencies already run: prebuilt connectors for M365, Google Workspace, ServiceNow, Salesforce, SAP, Oracle, file shares, email, SMS, and legacy databases, plus the Records API and an ETL lane for bulk migrations. Each record is normalized, hashed, and handed to the vault with its provenance intact.

See Integrations & capture

Records vault

Single-tenant store · Lifecycle · Holds

A single-tenant store built for federal records. Every record carries a NARA-aligned metadata profile with room for agency-specific fields. Lifecycle services run retention schedules, legal holds, declassification queues, and disposition. The vault is the platform’s source of truth: the substrate every other layer reads from.

AI orchestration

Prompt Runner · Model-agnostic

Prompt Runner executes AI operations across the vault at scale. Predefined prompts run continuously against new records or on demand against historical sets. The layer is model-agnostic: it routes inference to whichever model fits the workload and the classification level.

See AI in Feith

Workflow engine

Low-code · Routing · SLA

A low-code engine that drives any agency process. Workflows trigger on capture events, AI outputs, schedule changes, or user actions. Routing, approvals, escalation chains, and SLA enforcement are all configured visually, with no code deployments.

See Workflow & automation

Access layer

Apps · APIs · MCP · Portals

How users and systems reach records: the Feith applications (RMA iQ, FOIA Workbench, Record Insight, Case Manager), public-facing portals, the Records API, and the Feith MCP server. Every path enforces the same role and clearance checks.

Inside the boundary

Identity, audit, and compliance wrap all five layers. Every layer reads from one identity provider, writes to one audit log, and inherits one compliance posture.

Deployment topology

The physical view, generalized.

Where the logical stack runs in real infrastructure. This is the reference deployment: the same topology runs in commercial cloud, agency cloud, and on-premises datacenters. Component selection follows your standards.

Edge External services, outside the security boundary
Internet
DNS
Load balancer
Identity provider
Secure remote access
Cloud or datacenter container Your cloud subscription or on-prem datacenter
Single-tenant private subnet Inside the boundary
Web application server L5 access ingress
Application server L2–L4 platform logic
Full-text search server Index over the vault
Storage layer Inside the container, outside the subnet
Relational database
Object storage
Shared file system
Mail service Outbound notification routing, inside the container

Component selection follows your standards. On AWS GovCloud: DNS via Route 53, load balancing via ELB, RDBMS via RDS, object store via S3, shared file system via EFS. On Azure Government, the native equivalents. On-premises, your existing standards: F5, Oracle, NetApp, and the rest. Same topology, same code, same APIs.

Single-tenant architecture

Your records. Your subnet.

Every Feith deployment is single-tenant: no shared compute, no shared database, no shared storage with other agencies or tenants. Your records live in your subnet, inside your boundary, in infrastructure that exists only to serve your mission.

Not this

Shared-tenant

Agency AAgency BYouAgency D
Shared database Shared storage
  • Co-mingled compute
  • Shared database, shared storage
  • Noisy-neighbor performance
  • Vendor-imposed maintenance windows
Feith architecture

Single-tenant

Your agency
Your database Your storage
  • Infrastructure-level isolation
  • Dedicated database, dedicated storage
  • Predictable, dedicated performance
  • Maintenance on your schedule

AI inference happens against your records inside that boundary, and results flow back into your vault. Workflow execution stays in the same envelope. Your data, processing, and metadata never co-mingle with anyone else’s, including Feith’s other customers.

Deployment options

Deploy where the work lives.

Three options, equal capability. The choice is about where the records sit and who manages the infrastructure underneath.

On-premises

Where it runs
Your datacenter, your hardware
Who manages what
You manage the infrastructure. Feith manages the platform layer.
Environments
NIPR · SIPR · JWICS
Time to deploy
8 to 16 weeks, typical
HA / DR
Configurable to your RPO and RTO

The full platform inside your datacenter, behind your firewall, wired into your identity, monitoring, and backup. Required for many classified deployments. Full agency control of every layer of infrastructure.

Agency cloud

Where it runs
Your cloud subscription (AWS GovCloud, Azure Government, AWS TS C2S, others)
Who manages what
You manage the cloud account and identity. Feith manages the platform layer.
Environments
AWS GovCloud · Azure Gov · AWS TS C2S
Time to deploy
6 to 12 weeks, typical
HA / DR
Configurable to your RPO and RTO

A single-tenant deployment inside the agency’s own cloud subscription, running in compute you already pay for and accredit, with native integration to your cloud identity, key management, and logging.

Feith FedRAMP cloud

Where it runs
Feith’s FedRAMP Moderate (High IIP) authorized cloud
Who manages what
Feith manages infrastructure and platform. You manage records, users, and workflows.
Environments
FedRAMP Moderate (High IIP)
Time to deploy
4 to 8 weeks, typical
HA / DR
Multi-AZ active/passive standard. Active/active and multi-region available.

The fastest path for civilian agencies with FedRAMP Moderate requirements. Feith carries the authorization and the operational burden, so your team focuses on records and workflow, not infrastructure.

All three run the same code, the same APIs, and the same UI. Capability is identical across environments; the choice is about where your records live, not what the platform can do.

Operating environments

Eight environments. One platform.

Feith is in production across the full federal classification spectrum. Each environment maps to one or more deployment options.

Environment Classification Available on
NIPRNet Unclassified
On-premAgency cloudFedRAMP cloud
SIPRNet Secret
On-premAgency cloud
JWICS Top Secret / SCI
On-premAgency cloud
AWS GovCloud (Moderate) CUI
Agency cloudFedRAMP cloud
AWS GovCloud (High) IL5
Agency cloud
AWS Top Secret C2S Top Secret
Agency cloud
Azure Government CUI / IL5
Agency cloud
On-premises datacenter Any
On-prem

Availability reflects current deployments. Additional environments are supported on request wherever an accreditation path exists.

Identity & access

Identity runs through every layer.

Identity isn’t a layer; it’s a cross-cutting service inside every layer above it. The same provider authorizes a connector capture, a vault read, an AI inference, a workflow approval, and a Records API call.

Federation

SAML 2.0OIDCOAuth 2.0

SAML 2.0, OIDC, and OAuth 2.0 throughout. Single sign-on with your existing identity provider. Service accounts authenticate through the same federation patterns as interactive users.

Directory integration

Active DirectoryEntra IDLDAP

Active Directory, Microsoft Entra ID, and LDAP. Group membership drives role assignment; provisioning and deprovisioning flow through the directory, so a user’s access changes the moment their directory record does. No separate Feith user store to keep in sync.

Federal PKI

PIVCACPIV-I

PIV, CAC, and PIV-Interoperable credentials. Smart-card authentication for interactive sessions; certificate-based authentication for service accounts and machine-to-machine integrations. Compatible with the federal PKI common policy framework.

Multi-factor and cryptography

MFAFIPS 140-2/3TLS 1.3AES-256

MFA enforced on every access path. FIPS 140-2 and 140-3 validated cryptographic modules throughout. TLS 1.3 in transit, AES-256 at rest. Key management integrates with your KMS or HSM where required.

Access control and audit

RBAC + ABACTamper-evident logSIEM forwarding

Role-based access control with attribute extensions for clearance level, compartment, and need-to-know. Every access, action, AI inference, and administrative change writes to a tamper-evident audit log that feeds your SIEM over standard syslog or API ingestion.

AI model deployment

Any model. Your boundary.

The platform is model-agnostic. The orchestration layer routes inference to the model that fits the workload and the classification level. Records never leave the boundary, and they never train external models.

Cloud-hosted Anthropic · OpenAI (Azure) · AWS Bedrock · Google Vertex
Open-weight Llama-family · Mistral · others (swappable)
Unclassified
Routed
Routed
CUI · IL4
Routed, agency policy
Routed
IL5
Restricted
On accredited cloud
Secret / IL6
Not available
On-prem or enclave
Top Secret · SCI
Not available
Air-gapped

Records never cross the boundary: the model comes to the records. Swap models in orchestration config; no re-architecture.

Unclassified

For unclassified workloads, the platform connects to cloud-hosted models from major providers. Inference requests carry your records only as transient context, with no training, no logging, no retention beyond the call.

Classified · air-gapped

For classified and air-gapped workloads, the platform runs open-weight models on-premises or inside accredited cloud enclaves. They operate entirely inside your security boundary, with no external connectivity required.

Hybrid

Hybrid deployments route each workload to the right model by classification, sensitivity, and performance. As models improve, you swap them in orchestration, with no re-architecting workflows or applications.

Book a demo or an architect call

See it in your environment.

A demo shows the platform end to end. An architect call goes deep on topology, identity, models, and accreditation paths for your specific environment.

Request a session

Architect calls run 60–90 minutes with Feith engineering. NDAs are signed in advance when classified deployment is on the table.