Base URL
Your one change
Your SDK sends requests to the gateway instead of the provider. The key in the URL identifies your org, and the agent in the URL picks the policy.
// Product / Intent Gateway
No SDK. One line of config.
Check every tool call before your agent runs it.
Intent Gateway sits between your agent and its model provider. Point your agent at it with one base URL, and every tool call the model returns is checked against your policy before your code sees it.
// The Problem
The model picks the tool. Your agent runs it, no questions asked.
A tool call arrives inside the model's reply, and agent code executes it right away. Putting a check on that step usually means adding an SDK to every agent and an integration for every framework. Any agent that never got that integration runs with no policy at all.
// Without a checkpoint at the model API
- 01 /A tool call runs the moment the model asks for it
- 02 /Every agent framework needs its own integration
- 03 /Agents nobody instrumented run with no policy
- 04 /Nothing shows which tools each agent asked for
Setup
One line in the agent you already run.
Keep your framework, your prompts, your tools and your provider key. Change where the SDK sends requests, and ArmorIQ sits in the path.
Create a gateway key
Open Gateway in the ArmorIQ dashboard and create a key. It lives in the URL, so there is nothing new to load in your code.
Choose the agent
Pick the agent whose policy should apply. Leave it out to watch traffic first, and add enforcement when you're ready.
Ship the line
Paste it into your client or your environment. The dashboard confirms the first request the moment it arrives.
// Capabilities
What that one line gets you.
From the first request, without touching the rest of your agent.
01// See
Every request on one dashboard
Requests, tool calls and token use land in ArmorIQ under the agent that made them, whatever framework it runs on.
02// Enforce
Your policy on every tool call
Allowed calls pass through untouched. Blocked calls never reach your code, and the reply tells the agent which rule stopped them.
03// Capture
A signed plan for every turn
Each model reply that asks for tools is recorded as a plan with a signed intent token, the same evidence the ArmorIQ SDK produces.
04// Fail closed
Closed when in doubt
If ArmorIQ can't reach a decision for a named agent, the call is held back. An outage never turns into an open door.
Architecture
Three checks between your agent and the model.
All of it happens in the request path, on ArmorIQ's side. Your agent sees an ordinary model reply.
Plan
Every model turn
When a reply asks for tools, the gateway registers those calls with ArmorIQ as a plan and gets a signed intent token back.
Verdict
Every tool call
Each call is checked against the agent's active policy. Allowed calls go through as the model wrote them, and blocked ones are withheld with the reason.
Try it
Works with the SDK you already use.
OpenAI SDK
Chat completions, streaming and tool calls. Set OPENAI_BASE_URL.
Anthropic SDK
Messages, streaming and tool use. Set ANTHROPIC_BASE_URL.
Gemini
Use the OpenAI SDK with your Gemini key and a Gemini model.
Monitor mode
See what your policy would block while every call still runs.
Approvals
Approve a held tool call and let the agent's next try go through.
Something else
Another provider or framework? Tell us what you run.
Onboarding open
Live Intent Assurance↗Ready to control what your AI agents actually do?
Join the teams shipping safer, compliant AI agent deployments. White-glove onboarding for the first 50 design partners.