Portkey

AI Gateway for 1600+ LLMs — observability, guardrails, and governance in one platform

Portkey is an open-source production stack for GenAI teams. Unified API for 1600+ LLMs with observability, guardrails, prompt management, and AI governance. MCP Gateway centralizes authentication and access control for MCP servers across the org. Integrates in 3 lines of code. 3000+ teams, trillions of tokens processed.

Best for agent capabilities

Observability and evaluation

Useful in agent tasks

Research the web

Find, inspect, extract, and synthesize reliable information from the open web.

Build and deploy software

Write, execute, test, and ship code from an agent workflow.

Manage memory and knowledge

Keep useful context available across long-running tasks and sessions.

Automate business workflows

Connect services and coordinate durable multi-step operational work.

Overview

Websiteportkey.ai

Authapi key

PricingFree tier available · $0/month

SDKstypescript, python

OpenAPIYes

MCP SupportYes

When to use Portkey

Portkey is listed for Observability, Monitoring workflows. Portkey is an open-source production stack for GenAI teams. Unified API for 1600+ LLMs with observability, guardrails, prompt management, and AI governance. MCP Gateway centralizes authentication and access control for MCP servers across the org. Integrates in 3 lines of code. 3000+ teams, trillions of tokens processed. Consider it when an agent needs this capability through a documented service boundary instead of custom infrastructure maintained inside the application.

In 4agent's task model, the product maps to observability and evaluation. That makes it a candidate for workflows that research the web, build and deploy software, manage memory and knowledge, automate business workflows.

Integration checklist

Access: confirm the api key authentication flow, credential scope, rotation process, and whether calls can be attributed to a user or service identity.
Client support: official or documented clients are listed for typescript, python.
Automation: OpenAPI is listed as available, MCP is listed as available, and CLI support is listed as available. Confirm each requirement in the provider's current documentation.
Operations: test rate limits, timeouts, retry safety, error payloads, logs, data retention, and the current free-tier and paid pricing model with production-like volume.

Sources and verification

Use these primary sources to verify current behavior before making an implementation or purchasing decision. Product capabilities, limits, and prices can change after a 4agent listing is reviewed.