Portkey
Unified API to 1,600+ LLMs with gateway, guardrails, and governance
What is Portkey?
Portkey is a production stack for GenAI builders that combines an AI gateway, observability, guardrails, governance, and prompt management in one platform. It exposes a unified API to 1,600+ LLMs so teams can switch or fall back between providers without rewriting client code, and it advertises 99.9% gateway uptime. The open-source self-hosted gateway is genuinely free with no request cap, which is rare in this category.
Knowledge bases, internal search, operations, data, finance, HR, and back-office tools with AI workflows.
See the full Knowledge & Ops guide to compare more tools, buyer criteria, and related workflows.
Use cases to evaluate
Fall back from OpenAI to Anthropic to a local model when a provider has an outage
Apply PII redaction and prompt-injection guardrails to every outbound LLM call
Enforce per-team monthly token budgets and rate limits via service-account API keys
Cache repeated prompts at the gateway to cut bill and tail latency
Fit to evaluate
Platform teams running LLM infrastructure for many internal product teams
Multi-model shops that route by cost, latency, or capability per request
Enterprises needing SOC2 Type 2, HIPAA, and VPC deployment
OSS-first teams wanting the self-hosted gateway with no per-request fee
Business fit
Right for you if you need multi-provider routing, fallbacks, PII guardrails, and per-team budgets in one layer instead of stitching three vendors together. Skip if you're committed to one model provider and don't want the latency of another proxy in front of it. The free Developer tier gives 10K logs/month, but the production sweet spot is the $49/month plan with 100K logs and overages at $9 per extra 100K. Enterprise self-hosted into a private VPC is the path for regulated buyers.
How to evaluate Portkey
Use this category when operational data, policies, tasks, or internal requests are spread across disconnected systems.
Confirm the exact workflow
Map Portkey to one concrete workflow first, such as fall back from openai to anthropic to a local model when a provider has an outage. Avoid buying before the owner, trigger, output, and success metric are clear.
Check category fit
Compare internal search, permissions, workflow support, and reporting.
Compare practical alternatives
Shortlist Portkey against Glean, Guru, Slite so the decision is based on fit, effort, and workflow ownership rather than brand recognition alone.
Validate cost and rollout effort
Self-hosted open source is free and uncapped; Developer free with 10K logs/month, 3-day log retention, 3 prompt templates; Production $49/month covers 100K logs, $9 per extra 100K up to 3M, 30-day log retention, RBAC; Enterprise custom for 10M+ logs/month with SSO, VPC, SOC2/HIPAA, private cloud. Also confirm implementation time, support needs, and whether the medium setup matches your team.
Compare Portkey with alternatives
Use this quick comparison before booking demos or moving data into a new system.
| Primary workflow | Fall back from OpenAI to Anthropic to a local model when a provider has an outage, Apply PII redaction and prompt-injection guardrails to every outbound LLM call |
|---|---|
| Best-fit team | Platform teams running LLM infrastructure for many internal product teams, Multi-model shops that route by cost, latency, or capability per request |
| Implementation effort | Medium setup and maintenance profile |
| Pricing check | Free plan + paid plans |
| Closest alternatives | GleanGuruSliteSlab |
Portkey pricing
| Model | Free plan + paid plans |
|---|---|
| Snapshot | Self-hosted open source is free and uncapped; Developer free with 10K logs/month, 3-day log retention, 3 prompt templates; Production $49/month covers 100K logs, $9 per extra 100K up to 3M, 30-day log retention, RBAC; Enterprise custom for 10M+ logs/month with SSO, VPC, SOC2/HIPAA, private cloud. |
| Checked |
Common questions about Portkey
What is Portkey?
Portkey is a production stack for GenAI builders that combines an AI gateway, observability, guardrails, governance, and prompt management in one platform. It exposes a unified API to 1,600+ LLMs so teams can switch or fall back between providers without rewriting client code, and it advertises 99.9% gateway uptime. The open-source self-hosted gateway is genuinely free with no request cap, which is rare in this category.
What is Portkey used for?
Common use cases: Fall back from OpenAI to Anthropic to a local model when a provider has an outage; Apply PII redaction and prompt-injection guardrails to every outbound LLM call; Enforce per-team monthly token budgets and rate limits via service-account API keys; Cache repeated prompts at the gateway to cut bill and tail latency.
How much does Portkey cost?
Self-hosted open source is free and uncapped; Developer free with 10K logs/month, 3-day log retention, 3 prompt templates; Production $49/month covers 100K logs, $9 per extra 100K up to 3M, 30-day log retention, RBAC; Enterprise custom for 10M+ logs/month with SSO, VPC, SOC2/HIPAA, private cloud.
Who is Portkey best for?
Portkey fits Platform teams running LLM infrastructure for many internal product teams, Multi-model shops that route by cost, latency, or capability per request, Enterprises needing SOC2 Type 2, HIPAA, and VPC deployment, OSS-first teams wanting the self-hosted gateway with no per-request fee. Right for you if you need multi-provider routing, fallbacks, PII guardrails, and per-team budgets in one layer instead of stitching three vendors together. Skip if you're committed to one model provider and don't want the latency of another proxy in front of it. The free Developer tier gives 10K logs/month, but the production sweet spot is the $49/month plan with 100K logs and overages at $9 per extra 100K. Enterprise self-hosted into a private VPC is the path for regulated buyers.
What are alternatives to Portkey?
Common alternatives to Portkey include Glean, Guru, Slite, Slab, Tettra, Sana.