Portkey

Agents authenticate once to Portkey's MCP Gateway, which sits between them and MCP servers: for each tool call the gateway checks that user's access to the server and tool, then injects the upstream credentials (OAuth, API keys or custom headers, with external identity providers such as Okta and Auth0), with servers and individual tools provisioned per workspace and user. Agents therefore take authenticated action in outside systems through the gateway.

Sourcedocs.portkey.ai/docs/product/mcp-gatewayread 2026-09-22

The Agent Gateway is a proxy between agent clients and the customer's own agent servers, handling authentication, access control and observability for registered A2A agents; the agents and their steps run on the customer's servers. Portkey documents no workflow definition or agent execution of its own, so nothing in it sequences or branches agent steps.

Sourcedocs.portkey.ai/docs/product/agent-gatewayread 2026-09-22

Embedding and file requests route through Portkey to providers and responses are cached, but no retrieval structure that Portkey maintains over the customer's documents is documented; routing embedding calls builds none.

Sourceportkey.ai/pricingread 2026-09-22

Inputs and outputs pass through Portkey Guardrails, which act on the verdict automatically, denying the request, flagging it and letting it through, retrying or falling back, and the MCP Gateway applies content filters per server, per user tool provisioning and rate limits on tool calls; org and workspace admins can enforce guardrails, budgets and rate limits. These are customer controlled constraints on what an agent may do. No step where a person reviews, approves or must validate before an agent action commits is documented.

Sourcedocs.portkey.ai/docs/product/guardrailsread 2026-09-22

Security @ Portkey documents SSO through OIDC on enterprise plans, fine-grained role-based access control, encrypted storage of provider keys, TLS 1.2+ in transit and AES-256 at rest, and states compliance with SOC 2, ISO 27001, GDPR and HIPAA with regular third-party audits and penetration tests, linking the Trust Center at trust.portkey.ai; the pricing page names SOC 2 Type 2 and custom BAAs on Enterprise, and the docs cover SCIM for Okta and Microsoft Entra, audit logs, JWT authentication and API key rotation.

Sourcedocs.portkey.ai/docs/product/enterprise-offering/security-portkeyread 2026-09-22

Portkey records what agents did through it: the MCP Gateway logs every tool call with the user, parameters and response, filterable by server, user or time range, and model requests carry logs, traces, feedback, custom metadata, filters and alerts on every plan (pricing page), with retention stated per plan. Enterprise adds audit logs, logs export and OpenTelemetry export.

Sourcedocs.portkey.ai/docs/product/mcp-gatewayread 2026-09-22

The documented features are routing, caching, observability, guardrails, prompt management and the MCP and agent gateways. No memory layer with a stated scope and lifetime that persists an agent's state or a user's context across sessions is documented, and a response cache is not one.

Sourceportkey.ai/pricingread 2026-09-22

The pricing page lists private cloud deployment, VPC hosting and data isolation on Enterprise, and an open source gateway the customer hosts itself; the docs cover hybrid deployments on the customer's AWS (EKS, ECS, Marketplace), GCP and Azure (AKS, ACA), air gapped deployment, and a Gateway URLs setting that points the dashboard at the customer's own self hosted AI and MCP gateways. Customers can deploy into an environment they name, though some options sit on Enterprise.

Sourceportkey.ai/pricingread 2026-09-22

Nothing ready made ships for a buyer to adopt, whether agents, workflows, templates or role specific agents: the Prompt Library holds the customer's own templates, the Skills Registry shares skills the team writes, and the prebuilt assets are guardrail checks and partner guardrail integrations, which constrain agents rather than do work.

Sourcedocs.portkey.ai/docs/product/guardrailsread 2026-09-22

Work reaches agents through Portkey when a client sends a request: model, MCP and A2A calls are all made by the caller, and the Agent Gateway proxies them to the customer's agent servers. No schedule, event, webhook or inbound queue that starts an agent run is documented. As a gateway that waits for callers, it never wakes an agent on its own.

Sourcedocs.portkey.ai/docs/product/agent-gatewayread 2026-09-22

Model choice sits with the customer: Portkey's Universal API puts many model providers behind one interface, and routing configs the customer writes set providers, model overrides, fallbacks, load balancing weights, retries, conditional routing and circuit breakers, with support for private model deployments on AWS, GCP and Azure.

Sourceportkey.ai/pricingread 2026-09-22

Developers get Portkey's own API and SDKs: the Feedback page shows the same call through the Node and Python SDKs and the REST API, the docs index links a published OpenAPI specification, prompts are served through a Prompt API, and the Agent Gateway acts as a server to downstream agent clients for registered A2A agents.

Sourcedocs.portkey.ai/docs/product/observability/feedbackread 2026-09-22

Canary testing sends a set share of production traffic to a new model or prompt through a routing config, on every plan, and each request carries guardrail verdicts (pass or fail per check, counted in the log drawer) and weighted feedback scores sent through the Feedback API, analyzable by metadata. Guardrails can deny, retry or fall back on a failed check. Together that gives a documented way to evaluate a change with readable, comparable results.

Sourcedocs.portkey.ai/docs/product/ai-gateway/canary-testingread 2026-09-22

Portkey documents no browser, desktop or remote computer session that it runs for an agent; its gateway routes model calls, including provider computer-use tools whose actions the customer's own code carries out. It wires in no headless fetch or third party browser engine either.

Sourceportkey.ai/pricingread 2026-09-22