Skip to main content

DVARA vs Portkey

From observability to governance.

Portkey offers solid observability and a unified API across 1600+ models. DVARA is governance-first across 14+ providers (OpenAI, Anthropic, Gemini, Bedrock, Mistral, Cohere, Groq, Azure OpenAI, Ollama, Qwen, DeepSeek, Moonshot, ChatGLM, Grok), and the provider count is not the model count: /v1/models queries each configured provider live and returns the union of their catalogues — 84 chat models on a stock gateway with only OpenAI and Anthropic configured (measured, August 2026). With Policy-as-Code, tamper-evident audit trails, MCP tool call governance, and enterprise compliance.

Where Portkey is ahead

Read from Portkey's own documentation on 28 August 2026. A comparison in which we win everything is a marketing asset, not a comparison — so this is the part we would rather you knew before choosing.

  • Over 1,600 models across 40+ providers, and 60+ prebuilt named guardrails you switch on — we have the detection engines but no catalogue of one-click guardrails.
  • Managed edge deployment with a 99.99% uptime SLA and a published 20–40 ms overhead. DVARA is self-hosted or single-region, and publishes no latency figure at all.
  • Virtual keys: one object carrying the provider binding, budget, rate limit and permissions. Ours are two separate objects, which is more to wire up.

Source: portkey.ai

DVARA leads or ties every row in this table Portkey does not come out ahead on any of them. That is unusual enough that you should check it rather than take it on trust: these rows are the capabilities we built the product around, so they are the ones we chose to compare, and a table chosen by us is not a neutral survey. Portkey is likely to be the better tool where its own strengths lie — read its documentation alongside this.

Core Gateway

FeatureDVARAPortkey
Multi-provider unified API
SSE streaming support
Structured outputs across providers
Capability-aware route filtering
Circuit breaker per provider
Canary routing + A/B testing

Governance & Compliance

Portkey has observability for engineers. DVARA has governance for compliance teams, CISOs, and regulated industries.

FeatureDVARAPortkey
Policy-as-Code engine (YAML DSL)
Policy dry-run before activation
Tamper-evident HMAC-signed audit trail
PII detection and redaction
SOC2 / HIPAA / GDPR evidence packages
EU data residency enforcement
Prompt firewall + jailbreak detection

MCP, Agentic & A2A Governance

Portkey ships an MCP Gateway. DVARA pairs an MCP gateway with per-tool policy, approval gates, agent loop detection, signed audit, and per-tool cost attribution — a fuller agentic governance surface.

FeatureDVARAPortkey
MCP tool calls proxied and governed
MCP PII scan on arguments + responses
MCP server registry + credential store
Human approval gate (enforced at execution)
Agent loop detection + auto-kill
Agent-to-agent (A2A) hop governance
Session timeline (LLM + MCP unified)

Deployment & Enterprise

FeatureDVARAPortkey
Self-hosted / on-prem deployment
Air-gapped deployment
SSO via OIDC / SAML 2.0
RBAC access control
Multi-region active-active
Kubernetes Helm chart

The Bottom Line

Portkey is a capable LLM gateway with governance features. DVARA differentiates on agentic depth — per-tool MCP policy, approval gates, signed-audit chain integrity — and on running in your own infrastructure under predictable annual licensing rather than a metered SaaS bill. For teams whose compliance posture leans on a tamper-evident audit chain or whose agent workloads need first-class tool-call governance, DVARA is the deeper option.

Ready to migrate? See the step-by-step guide.

Ready to see the difference?

Run the LLM Gateway free in your own infrastructure, production included — a licence when you need the MCP and A2A planes with support.

The Portkey column describes what Portkey's public documentation said when we read it in June 2026 — a dash means we did not find the capability documented, not that the product cannot do it. Vendors ship continuously; check their docs before deciding. The DVARA column is a claim about our own code and is verified against the released tag.