Setup
One credential change
In n8n, add an OpenAI credential that points at the gateway. LangGraph, CrewAI and Flowise take the same base URL.
Keys stay inside. The gateway adds the provider key, so your workflows never hold it.
Automation Agencies & Builders
You ship agents into someone else’s systems.
Your workflows run with your clients’ credentials. When a client asks what the agent can do, or why it did something, you need a better answer than a screenshot of the workflow.
In one sentence: StartiQOS secures n8n, LangGraph, CrewAI and Flowise agents with one base-URL change, keeps provider keys out of workflows, and gives each client an exposure report.
The Problem
Fast to build. Hard to vouch for.
The speed that wins the project also leaves the gaps.
01
Keys live inside workflows
Provider and API keys pasted into nodes travel with every export and every copy.
02
Broad access by default
It is faster to give an agent full access than to scope it, so most agents get it.
03
One bad document reaches the client
An email or file with a hidden instruction can steer an agent that holds the client’s access.
04
Clients ask for proof
Larger clients want to know what was allowed, what was blocked and who approved it.
What You Get
One base URL. A policy per agent.
StartiQOS sits in front of the model your workflow already calls.
Per client
A policy for each agent
The agent and task names travel in the base URL, so each client’s agents get their own learned, approved policy.
Deliverable
An exposure report
After two weeks in shadow mode, a report for each agent of the access it holds against the access it used.
Something to review with the client, not just a promise.
Fit
Is this the right time?
Early access works best for builders with agents already running for clients.
A good fit if
- You build on n8n, LangGraph, CrewAI or Flowise
- Your agents call an OpenAI-compatible model API
- Clients ask about security or data access
- You can run a container for each client
Not a fit yet if
- You need one gateway shared by many separate client organisations (coming next)
- Your agents act through MCP servers or desktop tools (coming soon)
Questions
What builders ask us.
Yes. Create an OpenAI credential in n8n with the gateway key and the gateway base URL. The setup script prints both. Your workflow does not change.
Yes. Each gateway is a container you can run in the client’s environment or your own. One gateway shared across separate client organisations is coming next.
Each check takes under 0.05 ms at the 99th percentile in our tests, far less than the model call itself.
The exposure report from the pilot: for each agent, the access it holds against the access it used, and every combination that could move data out.
Give clients an answer
they can check.
Two weeks of shadow mode, nothing blocked, and an exposure report for every agent at the end.