Banking and finance
Banking secrecy, model risk and operational resilience: AI under the same requirements as the rest of the information system.
Enterprise AI in production, secured and governed.
Navin AI Platform brings models, agents, tools and data together in a single Control Plane. Every call goes through a gateway, a policy and an audit trail, in the cloud of your choice or in your own data center.
- Kubernetes
- PostgreSQL
- OpenTelemetry
- Keycloak
- Open Policy Agent
- SPIFFE
- Cilium
- vLLM
See the 62 services of the Navin Manager catalog
Security
- ClamAV
- Falco
- Open Policy Agent
- OpenFGA
- Presidio
- Tetragon
- Tink
- YARA-X
Identity and secrets
- Keycloak
- LemonLDAP::NG (WebSSO)
- OpenBao
- SPIRE
Gateways
- Apache APISIX
- HAProxy
- LiteLLM gateway
- MCP hosting
AI and inference
- Docling Serve
- Haystack
- LiveKit Egress
- LiveKit Ingress
- LiveKit server
- LiveKit SIP
- Ollama
- vLLM
- Voice agent workers
Agent memory
- Cognee
- Mem0
Data
- ClickHouse
- etcd
- Label Studio
- Marquez
- MySQL Community Server
- NATS
- OpenLineage SDK
- OpenSearch
- OpenSearch Dashboards
- PostgreSQL (sandbox)
- Qdrant
- Redis Open Source
- Temporal
- Temporal Web UI
Evaluation
- garak
- LLaMA-Factory
- Opik
- promptfoo
- Ragas
Observability
- Alertmanager
- Blackbox exporter
- Grafana Alloy
- Jaeger
- Node exporter
- OpenTelemetry Collector
- PostgreSQL exporter
- Prometheus
- Redis exporter
Storage
- SeaweedFS object storage
Network
- Asterisk PBX
- Kamailio SIP edge
- keepalived
- RTPengine media relay
Navin platform
- Navin Enterprise
- Navin SM
Why Navin
Four guarantees, enforced on every call
Security, governance, traceability and scalability are not configured project by project: they sit on the path of every request.
01 - Security
Defense in depth, from the browser to the network: Coraza WAF and OWASP rules at the gateway, a default-deny Cilium network, tenant isolation by PostgreSQL RLS. Secrets stay in OpenBao and no agent ever sees them.
02 - Governance
OPA policies stacked from the platform down to the project, four-eyes approvals, a kill switch on any agent, tool or model. Rules are enforced by the gateways, not left to each team's code.
03 - Traceability
Every call carries a W3C traceparent and produces OpenTelemetry spans, a Usage Event and a hash-chained audit event kept in WORM storage. Any run can be replayed and proven.
04 - Scalability
HPA and KEDA autoscaling, NATS JetStream queues, durable checkpoints for agents, quotas and GPU pools per tenant. High availability is declared service by service.
The platform in figures
A complete platform, documented brick by brick
Every brick has its documentation, its API contract and the tests that prove its behavior.
- functional domains
- 8
- documented bricks
- 112
- deployment topologies
- 5
- official SDKs: Python, TypeScript, Java and Go
- 4
What you can do
The whole AI lifecycle, on a single platform
From models to agents, from data to the invoice: every domain has its gateway, its registries and its controls.
An agent runtime built for production
Agent loop, sub-agents, run plans and a message bus: the Navin runtime executes long-running, tool-using agents with checkpoints, sandboxing and replay. LangGraph and Temporal plug in through adapters.
- Durable execution and recovery after an incident
- gVisor or Kata sandbox for code and tools
- Kill switch and replay of every run
- Python, TypeScript, Java and Go SDKs
Related bricks
- C01Agent runtime
- C02Durable execution
- C11Human in the loop
- C12Execution sandbox
- C14Agent kill switch
- C15Execution replay
Deployment
One platform, five deployment topologies
The same controls apply everywhere, from a single component to a cluster cut off from the Internet.
01 - Single brick
The Control Plane, PostgreSQL and one component, for example the model gateway.
02 - Shared SaaS
One cluster, several tenants isolated by RLS, OpenFGA and Cilium namespaces.
03 - Dedicated tenant
A namespace or a cluster per key account, with its own OpenBao mount and GPU pool.
04 - Sovereign on premises
A cluster isolated from the network, local vLLM models, a Harbor mirror and an internal Git.
05 - Hybrid
Local vLLM for restricted data, remote providers allowed by policy for other classifications.
Solutions
Designed for the most demanding industries
Protect data, frame usage, prove what was done: shared needs, with answers shaped for each industry.
Insurance
Claims, underwriting and customer relations: agents that process documents and conversations, with a trace of every decision.Public sector
Sovereignty, on-premises hosting and reversibility: AI at the service of public servants and citizens.Healthcare
Health data, confidentiality and accountability: AI that assists care teams without exposing patients.Manufacturing
Know-how, technical documentation and maintenance: AI that protects industrial property.Telecommunications
Volumes, real time and B2B offers: a multi-tenant platform to serve your teams and your customers.
Resources
Latest publications
White papers, guides and datasheets to design and run enterprise AI.
- DatasheetAgents
Isolated MCP runtimes per user and project
The MCP registry and gateway decide who may use which server; the MCP runtime creates isolated instances on demand on the ToolHive Kubernetes Operator and destroys them after use. Every tool call still goes through policy, DLP and approval.
2 min read - DatasheetSecurity
Mandatory file scanning with ClamAV and YARA-X
Every file sent to the AI is quarantined, scanned by ClamAV and YARA-X and then validated before Docling, OCR or RAG can read it. Fully local, fail closed, logged and highly available.
2 min read - White paperGovernance
Governing AI agents in production
How Navin Enterprise AI keeps autonomous agents under control: OPA policies stacked from platform to tenant to project, approvals with a human in the loop, a kill switch available at any time and a governance registry as the single source of truth.
25 min read
Ready to put your AI agents into production?
A session with an architect, on your use cases and your constraints.