A unified way for AI agents to access MCP tools, without modifying your agents or servers. Self-host, deploy to your cloud, or run fully managed.
Built-in support for every auth type
Support for OAuth 2.1, API Tokens, Header Auth. Fully configurable
JWT validations, Identity forwarding, Passthrough Headers and other advanced auth features supported out of the box
Bring your own auth with Okta, Entra and more
Central registry for servers and tools
View available servers and tools in one place
Scope servers to specific teams, no sprawl
Integrate new MCP server in a few mins
Fine-grained access control for servers and tools
RBAC for org, workspace, team, or individual user, scope access precisely
No shadow tooling. If it's not approved in the gateway, agents can't use it
Let teams share tools across boundaries without sharing credentials
End-to-end observability for agent workflows
Get unified traces and see exactly which tools an agent called
Fast debugging
Every tool invocation is audited, logged, and searchable
Usage and performance analytics
Track request volume, latency, and errors across MCP traffic.
Identify slow or failing servers before they impact agents
Monitor usage patterns per tool, per server, per team
Runtime policy enforcement
Enforce PII redaction, content filtering, and compliance policies in one place
Stop unauthorized tool invocations before they execute
Consistent policies across LLMs and MCP
MCP Gateway integrates with the agent frameworks, identity providers, guardrails, and tools you already use.
Expose internal MCP tools without exposing credentials
Share secure, controlled access to your internal MCP servers. Authenticate users via Portkey, and manage permissions at the team level so credentials never leave the gateway.
Scale MCP access across your org
Give the entire org access to your MCP servers. Users authenticate with your existing IdP, no additional onboarding. All traffic routes through the gateway.
Bring vendor tools under your governance
Integrate vendor and partner MCP servers using OAuth, API keys, or custom auth. Bring them under the same governance and observability model as internal tools.
Connect, govern, and share MCP tools from multiple servers
across teams in your organization.
Secure by design
End-to-end encryption, SOC 2 compliance. Complete deployment flexibility—SaaS, private cloud, VPC, or fully self-hosted.
One control plane for LLMs and MCP
Integrate with Portkey's AI Gateway to manage your models and tools together. Unified observability and policy.
Open-source and community-driven
Built in the open with hundreds of contributors. Inspect the code, shape the roadmap, and deploy with confidence.
