B — Some risk signals Graded October 16, 2026Minor concerns worth a quick check.Based on public adoption and transparency signals as of October 16, 2026. Our opinion, not a code audit or security guarantee.
App connector · General
gateway
A blazing fast AI Gateway with integrated guardrails. Route to 1,600+ LLMs, 50+ AI Guardrails with 1 fast & friendly API.
Bottom line: Signal-based grade (4/7 points): very widely starred on GitHub, public source code. This reflects adoption and transparency, not a code audit.
- Popularity score
- 46/100
- Rank
- Not ranked: outside the 500 most popular tools
- GitHub stars
- 13.2k
What it does
Gateway is a unified API layer that routes requests to 1,600+ language models while enforcing 50+ built-in AI guardrails for safety and compliance. It abstracts away the complexity of managing multiple LLM providers and security policies through a single, high-performance endpoint.
Who it’s for
Engineering teams and enterprises needing centralized LLM access with integrated safety controls, and builders who want to avoid vendor lock-in while maintaining consistent guardrail enforcement across multiple models. It's particularly useful for organizations managing sensitive workloads across different AI providers.
Copy-paste setup
Check these steps against the project’s own instructions before running anything.
Verify against the repo README at https://github.com/Portkey-AI/gateway (opens external site) for current installation instructions. Likely installation via npm (npm install @portkey-ai/gateway) or Docker, with configuration through environment variables or API keys. Check the repository for MCP server integration details or REST API documentation.
Risks
- Credential scope exposure: Gateway handles API keys for 1,600+ LLM providers; compromise could leak credentials across multiple services
- Prompt injection at routing layer: Guardrails may be bypassed if routing logic is exploited or if guardrails don't cover all attack vectors
- Centralized data flow: All requests funnel through a single gateway, creating a single point of failure and potential data exfiltration vector
- Third-party dependency chains: Integrating 50+ guardrails and 1,600+ LLM endpoints multiplies supply-chain risk surface
- Guardrail effectiveness unverified: No transparent documentation of guardrail test coverage or independent security audits visible
How to stay secure
Least-privilege permissions
- Rotate API credentials regularly – Change passwords and access tokens for your LLM providers at least quarterly; treat this like changing locks on important doors
- Use separate credentials per environment – Keep development, testing, and production credentials completely isolated so a leak in one area doesn't compromise others
- Limit admin access – Only grant gateway administrator rights to 2-3 core team members; everyone else should have read-only access or no access at all
- Enable API key scoping – Request that your LLM providers restrict each API key to only the specific services your team actually uses, not all their offerings
Network isolation
- Deploy gateway behind a firewall – Place the gateway on an isolated internal network segment; only allow specific applications to communicate with it
- Use VPN for remote access – If team members access the gateway remotely, require them to connect through a VPN (virtual private network) first
- Disable unnecessary guardrails – Turn off any of the 50+ available guardrails that you don't actively need; every active feature increases risk
- Encrypt all connections – Ensure all data traveling to and from the gateway uses encryption (HTTPS/TLS), not plain-text communication
Monitoring
- Log all requests and access – Keep detailed records of who accessed what, when, and from where; review these logs weekly for unusual activity
- Set up alerts for anomalies – Create automated warnings if someone tries accessing the gateway at unusual times, from unusual locations, or requesting unusual volumes of data
- Track credential usage – Monitor which API keys are being used, when, and how often; unexpected spikes may indicate compromise
- Audit guardrail decisions – Periodically review cases where guardrails blocked requests to ensure they're working correctly and not being bypassed
Red flags to watch for
- Requests to unfamiliar LLM providers – If your logs show API calls to providers your team doesn't use, someone may have gained unauthorized access
- Spike in data volume – A sudden increase in requests through the gateway could indicate data exfiltration or a compromised credential
- Guardrails frequently disabled or bypassed – If users report that safety guardrails are interfering with legitimate work, investigate whether they're being circumvented
- Vendor security incidents – Monitor announcements from the gateway provider and all 1,600+ LLM providers; if any suffer a breach, immediately rotate your credentials with them
Alternatives
- Anthropic Claude API (direct, first-party guardrails but single-model)
- LiteLLM (lightweight LLM routing without integrated guardrails)
- Langchain / LlamaIndex (agent orchestration with guardrails plugin ecosystem)
- Hugging Face Inference API (multi-model access with content filtering)