Portkey: AI Gateway + Guardrails

Portkey is an AI gateway that combines model routing, observability, and guardrails into a single control plane. It sits between your agents and the LLMs they call, giving you policy enforcement, cost control, and security controls without rewriting your application.

• You need guardrails around prompt injection, PII leakage, or toxic outputs.
• Your organization requires approval workflows or budget caps for LLM usage.
• You run multiple agents or teams and want centralized model access policies.
• You want semantic caching and intelligent routing to reduce costs.

• Prompt and content guardrails. Detect injection, jailbreaks, PII, and unsafe content before it reaches or leaves the model.
• Centralized model access. One API key layer for all providers and teams.
• Semantic caching. Cache similar prompts to reduce redundant spend.
• Built-in observability. Traces, cost dashboards, and latency breakdowns.
• Enterprise controls. SSO, audit logs, RBAC, and compliance features.

• Guardrails are not perfect. Sophisticated attacks can still slip through; treat them as a layer, not a guarantee.
• Added latency. Every request passes through the gateway, plus policy checks.
• Paid features add up. Advanced guardrails and enterprise controls are not on the free tier.
• Provider dependency. You still need underlying LLM provider keys.

• Free tier for small projects and individuals.
• Paid plans start around $49/month and scale with requests and features.
• Enterprise pricing is custom and includes SSO, audit, and dedicated support.

Teams shipping agents in regulated or multi-user environments where control, visibility, and safety matter as much as model quality. Complements LiteLLM or can replace it if guardrails are the priority.

• OpenAI / Anthropic / Azure providers through the gateway.
• Langfuse for deeper tracing where Portkey observability is not enough.
• OpenClaw / Hermes agents via the OpenAI-compatible endpoint.