Skip to main content
30-Day Free Trial · No Credit Card Required

Govern, audit, and route every LLM and MCP call.

DVARA is an AI governance platform for LLM and MCP traffic — Policy-as-Code, immutable audit, PII + injection guardrails, agentic governance, and cost attribution in one gateway. Drop-in OpenAI compatibility across 14 providers, from $59/mo flat.

dvara-gatewayzsh
$docker run -p 8080:8080 -e OPENAI_API_KEY=sk-... \
> ghcr.io/dvarahq/dvara/dvara-llm-gateway:latest
DVARA Gateway started on :8080
LLM Proxy ready (MCP Proxy ships as dvara-mcp-gateway)
$ curl -X POST localhost:8080/v1/chat/completions \
> -H "Authorization: Bearer $TOKEN" \
> -d '{"model":"gpt-4o","messages":[...]}'
policy: PASS pii-scan: CLEAN audit: SIGNED cost: $0.0032
{"choices":[{"message":{"content":"..."}}],"usage":{"total_tokens":214}}
10K+
Requests/sec on 4 cores
<10ms
Added latency P95
14
Provider adapters
<2min
Zero to first request
Capabilities dashboard illustration

Everything You Need to Govern AI at Scale

One SKU. Every governance capability in the box from day one — from your first API call to enterprise audit and compliance.

🛡
Policy-as-Code Engine

YAML DSL for model allowlists, token limits, MCP tool restrictions, budget-based downgrades, and time-of-day rules. Dry-run mode before activation. Draft → Active → Shadow lifecycle with versioning and rollback.

🚨
Guardrails & Safety

In-process PII detection (regex always-on + Phileas opt-in with 17 filter types), prompt-injection and jailbreak detection, content filters, output-schema enforcement, and grounding/hallucination checks. OWASP LLM05 / LLM07 / LLM10 patterns. Optional ML classifiers (Lakera, Shield Gemini) and HTTP plugins.

🔏
Immutable Audit Trail

Every request HMAC-SHA256 signed with hash-chain integrity verification. Tamper detection built in. SOC2, HIPAA, and GDPR evidence packages generated on demand.

🤖
Agentic AI + MCP Proxy

MCP server registry and governed tool-call proxy with PII scanning on arguments and responses, human approval gates, and credential centralisation. Agent loop detection with auto-kill, session tracking across LLM and MCP calls, full session timeline.

🔐
Multi-Tenant + BYOK

Row-level tenant isolation on every query. Each tenant brings their own provider keys (AES-256-GCM at rest or vault-reference). Strict-BYOK rejects platform fallback when enabled. 3-level config hierarchy: global → tenant → API key.

💰
FinOps & Cost Control

Real-time per-request cost calculation. Budget caps per tenant, team, or API key. Automatic model downgrade at soft limit. Chargeback reports and cost forecasting.

🔄
Multi-Provider Unified API

Drop-in OpenAI-compatible endpoint. Route to OpenAI, Anthropic, Gemini, Azure, Bedrock, Ollama and 8 more with zero code changes. Streaming, structured outputs, and vision built-in.

Intelligent Routing

Round-robin, weighted, capability-aware, latency-aware, and cost-aware routing. Circuit breakers, failover, and retry with exponential backoff. Canary and A/B testing.

🔍
Full Observability

Prometheus metrics, OpenTelemetry distributed tracing, structured JSON logging. Reference Grafana dashboards in dvara-examples. Unified traces spanning LLM turns and MCP tool calls.

Route planning illustration

Two Governed Data Planes. One Control Plane.

The LLM Gateway sits at the edge for model traffic. The MCP Proxy sits inside your perimeter for tool calls. Both share the same policies, audit trail, and API keys.

LLM Data Plane
</>
Your App
Agent / API Client
{}
LLM Gateway
:8080
AI
Providers
OpenAI · Anthropic · Gemini
Control Plane
PoliciesAuthAuditBudgetsPII
MCP Data Plane
</>
Your App
Agent / IDE
{}
MCP Proxy
:8070
DB
Tools
DB · FS · Slack · APIs
Performance illustration

Built for Production Scale

Governance with zero performance tax. Every layer is optimised for throughput — lightweight concurrency, tamper-evident audit, multi-layer caching, and zero-copy streaming.

10,000+RPS
per node on 4 cores
Lightweight concurrency model handles thousands of simultaneous requests with minimal memory overhead.
<10ms
added gateway latency
Policy checks run in single-digit milliseconds. Governance adds near-zero overhead to your AI calls.
Tamper-evident
audit chain
Every event HMAC-SHA256 signed and hash-chained on the response path. Chain continuity is verifiable end-to-end — any gap or alteration is detectable.
Active-active
multi-region
Run multiple regions active-active with automatic failover. Data residency is preserved during failover — EU tenant traffic stays in the EU.
Lightweight Concurrency
Handles 10,000+ concurrent connections per node with minimal memory per request. No thread pool tuning required.
Zero-Copy Streaming
SSE tokens flushed immediately with no buffering or rewriting. First token arrives the instant the provider emits it.
Tamper-Evident Audit
Every event HMAC-SHA256 signed and hash-chained on the response path. Append-only by application invariant; chain continuity verifiable end-to-end.
Multi-Layer Caching
Hazelcast IMaps deliver sub-millisecond policy, auth, and config lookups. Cache-aside with PG NOTIFY invalidation on every mutation — fleet-wide consistency in seconds, not minutes.
Automatic Failover
Per-provider health monitoring with instant failover on degradation. Requests reroute seamlessly — your users never notice.
Stateless Data Plane
Restart any pod without coordination — Postgres holds all durable state; the Hazelcast cluster carries the replicated hot-path cache. The 2-connection HikariCP pattern keeps each pod lightweight; scale horizontally behind your HPA.

Governance Is the Architecture, Not a Feature

In every competitor, governance is bolted onto a proxy. In DVARA, governance is the design constraint that shaped every layer.

D1
Policy-as-Code with Dry-Run
Competitors: config flags or basic allowlists.

Full YAML DSL with version control, conflict detection at load time, and dry-run mode that evaluates the policy against a provided context before activation. Your auditors get proof that policies were tested before deployment.

D2
Immutable, HMAC-Signed Audit
Competitors: log to stdout or a mutable database.

Every record HMAC-SHA256 signed and hash-chained on the response path. Chain continuity is verifiable end-to-end — any gap or alteration is detectable. Append-only by application invariant, not by hopeful convention.

D3
MCP Proxy — No Competitor Has This
Competitors: governance stops at the LLM boundary.

Separately deployed, centrally governed MCP Proxy. Every tool call audited, policy-checked, and PII-scanned. Credential centralisation. Human approval gates enforced at execution.

D4
Agentic AI Governance
Competitors: nothing meaningful for agent loops.

Full OpenTelemetry trace spanning LLM turns + MCP calls. Agent loop detection with auto-kill. Human-in-the-loop approval for high-risk tool calls. Session-level cost and compliance summary.

Compliance analysis illustration

Audit-Ready from Day One

Generate compliance evidence packages on demand. Every request is immutably logged, every policy decision recorded, every PII event tracked.

Compliance OverviewLive
SOC2
SOC 2 Type II
Ready
HIPAA
HIPAA
Ready
GDPR
GDPR
Ready
Every
Request Audited
Signed
& Hash-Chained
0
Tampered
24/7
Reports
Recent Audit Events
12:04:32LLM Requestpolicy: ALLOWpii: CLEANHMAC signed
12:04:31MCP Tool Callpolicy: ALLOWpii: REDACTEDHMAC signed
12:04:29LLM Requestpolicy: DENYpii: CLEANHMAC signed
12:04:28MCP Tool Callpolicy: ALLOWpii: CLEANHMAC signed
12:04:26LLM Requestpolicy: ALLOWpii: CLEANHMAC signed
12:04:24LLM Requestpolicy: ALLOWpii: REDACTEDHMAC signed
12:04:22MCP Tool Callpolicy: DENYpii: CLEANHMAC signed
12:04:20LLM Requestpolicy: ALLOWpii: CLEANHMAC signed
12:04:32LLM Requestpolicy: ALLOWpii: CLEANHMAC signed
12:04:31MCP Tool Callpolicy: ALLOWpii: REDACTEDHMAC signed
12:04:29LLM Requestpolicy: DENYpii: CLEANHMAC signed
12:04:28MCP Tool Callpolicy: ALLOWpii: CLEANHMAC signed
12:04:26LLM Requestpolicy: ALLOWpii: CLEANHMAC signed
12:04:24LLM Requestpolicy: ALLOWpii: REDACTEDHMAC signed
12:04:22MCP Tool Callpolicy: DENYpii: CLEANHMAC signed
12:04:20LLM Requestpolicy: ALLOWpii: CLEANHMAC signed
Immutable Audit Trail
HMAC-SHA256 signed records with hash-chain integrity verification. Tamper detection built in.
Scheduled Reports
Weekly or monthly compliance report delivery in PDF and JSON.
PII Detection Log
Every redaction event logged with action taken, pattern matched, and tenant context.
Policy Decision Records
Every request logs ALLOW/DENY with rule ID and reason. Full dry-run history.
Right to Erasure
GDPR tenant data purge pipeline with confirmation and audit trail.
Data Residency
Per-tenant region pinning. EU tenant traffic never leaves EU — enforced during failover.

How DVARA Compares

The structural gaps in every competitor are not feature gaps — they require architectural redesigns to close.

DVARALiteLLMPortkeyKong AICloudflareHeliconeAWS BedrockBifrostTrueFoundryDatabricks
Core Gateway
OpenAI-compatible unified API
SSE streaming (zero-copy pass-through)
Structured outputs on Anthropic + Gemini
Multi-provider (OpenAI, Anthropic, Gemini, Bedrock, Azure, Ollama)
Designed for 10K+ RPS per node
Routing & Resilience
Weighted routing + failover
Latency-aware routing (EWMA P95)
Cost-aware routing
Canary + A/B routing
Per-provider circuit breaker
Capability-aware route filtering
Policy & Governance
Policy-as-Code engine (YAML DSL)
Policy dry-run before activation
Policy versioning + rollback
RBAC access control
SSO (OIDC / SAML)
PII & Data Protection
PII detection + redaction (LLM + MCP)
DLP custom patterns per tenant
Reversible tokenisation
EU data residency guaranteed
Right to erasure (GDPR)
MCP Proxy & Tool Governance
MCP tool calls proxied & governed
MCP server registry + credential store
MCP argument-level policy rules
MCP PII scanning (args + response)
MCP rate limiting + circuit breaker
Agentic AI Governance
Human approval gate (enforced at exec)
Agent loop detection + kill switch
Multi-agent session tracking
Full OTel trace: LLM turns + MCP calls
Audit & Compliance
Immutable HMAC-signed audit log
SOC2 / HIPAA / GDPR evidence packages
SIEM export (Splunk, CloudWatch, Kafka)
Scheduled compliance reports (PDF + JSON)
FinOps & Cost Control
Real-time cost per request (USD)
Budget caps (soft + hard) per tenant
Auto model downgrade on budget threshold
Chargeback reports per tenant/team
Guardrails & Safety
Prompt firewall (pre + post filter pipeline)
Jailbreak detection
Output sanitization (XSS, SQLi, SSRF)
System prompt leakage detection
Content policy filters per tenant
Deployment & Infrastructure
Air-gapped / on-prem deployment
Multi-region active-active
Kubernetes Helm chart + HPA
Hot config propagation (< 5s)
Observability
Prometheus metrics + Grafana dashboards
OpenTelemetry distributed tracing
Per-tenant analytics
Anomaly detection for traffic patterns
Coding assistant illustration

Developers Love It. Compliance Requires It.

Every stakeholder gets exactly what they need from the same platform.

Developer
"I'm calling five LLM providers with five different SDKs and no fallback."

One endpoint. Drop-in compatibility. Route to any provider with automatic failover. Add streaming, rate limiting, and observability in minutes.

Platform Engineer
"12 teams calling LLMs directly. Agents hitting databases with no governance."

Two governed data planes, one control plane. LLM Gateway at the edge, MCP Proxy inside the perimeter. Adopt incrementally — add tool governance when agents go to production.

Compliance Officer
"Auditors want proof of what AI systems were used and that no PII was leaked."

Immutable audit trail with tamper-evident signatures. PII detected and redacted before reaching providers. Compliance evidence packages generated on demand.

CISO
"Developers are calling GPT-4 from their laptops. We have no idea what data is leaving."

Hard policy enforcement at the gateway. PII blocked before it leaves your network. Role-based access to authorised models only. Every request logged and auditable.

CFO / FinOps
"AI spend is $40K/month and we can't tell which team is spending what."

Real-time cost tracking per request. Budget caps that automatically enforce. Monthly chargeback reports. Smart caching reduces repeated token spend by up to 40%.

CTO
"We need the governance layer we'd build in 18 months — but we need it now."

Two governed data planes — AI at the edge, tools inside the perimeter — managed as one. Production-grade performance. Enterprise-ready from day one.

See It in Action.
Start Your Free Trial Today.

Full access for 30 days. No credit card. Deploy in under 2 minutes and see governance working on your first request.

Start Free Trial