One governed path to the models your teams use. Your policy decides.
Your teams and agents reach approved models, tools and business APIs through one endpoint. Reign Gateway checks who is calling, applies your policy and spend limits, and records each model call. Reign runs it for you, 24×7.
Six controls on every model call.
Set by your policy, applied at one endpoint, and recorded.
Identity
Every request is checked. Each person or agent reaches only the models and tools you approve.
Models
OpenAI, Anthropic, Gemini, Bedrock, Azure OpenAI or your own open-weight endpoint, on your own provider accounts and keys.
Tools
MCP tool servers and business APIs such as GitHub, GitLab and Jira. Each tool is off until it is granted to an identity.
Spend
Request, token and dollar limits.
Content
Guardrails block or mask sensitive text, and can hand off to a guardrail service you already run. Content capture is off by default.
Record
Caller, model, tokens, cost and guardrail decisions for each model call and each MCP tool call, in storage you own.
Your developers point their SDKs at one endpoint: OpenAI-compatible, Anthropic-compatible or MCP.
The boundary every model call crosses.
Plus any HTTPS business API you approve.
- Caller
- The identity that made the call, with its rights
- Policy
- The policy and limits that applied
- Model or tool
- Where the call went
- Decision
- Allowed, blocked or masked, by which guardrail
- Time and cost
- When it happened, in tokens and dollars
- Stored
- In storage you own
Sample record. Not a customer.
Three situations it is built for.
Calls you cannot reconstruct
An application or an agent reaches a model, and afterwards nobody can say who called, which rules applied or what was decided. The record shows the caller, the model or tool, the policy that applied and the decision, for every call through the gateway.
Agents somebody else built
Your engineers, or a vendor, have built agents, and each one reaches models its own way. One boundary and one set of rules covers all of them, whoever built the agent.
A rule on paper that nothing enforces
You have a written rule about what may be sent to a model. The gateway applies it at the point the call is made, instead of asserting it afterwards.
Start with one traffic path. The policy gets written by people who have seen what it has to allow. Book a 30-minute governance briefing →
Reign carries the work. You keep the choice.
| Reign Gateway | Hosted AI gateways | Cloud AI gateways | Open source, self-run | |
|---|---|---|---|---|
| Install, run and upgrade, 24×7 | Reignmonitored and patched | Vendorshared service | Vendorthe service; you build on it | You |
| Policy and guardrails kept current | Reignupdated by engineers | You | You | You |
| Who chooses the model | Youany model, your keys | Youtheir supported list | Vendorfavors their catalog | You |
| MCP tools and business APIs | Reigngoverned access | Vendortheir supported list | Youbuilt on their services | Youplugins you build |
| Where your data goes | Youapproved destinations | Vendorthrough their service | Youyour cloud account | You |
| Evidence you own, every model call | Reignin your storage | Vendorlogs in their console | Youlogging you assemble | Youyou configure and keep it |
Category comparison of approaches, not a claim about any named product.
Open by design. Your models, your data, your choice.
Use frontier, your own or open-weight models, on your own accounts and keys. A gateway dedicated to you checks and records every model call, and you can swap models without rewriting your policy.
Open-source base
Built on the open-source agentgateway.
Operated for you
Reign installs, runs and upgrades the gateway, 24×7, and keeps your policies and guardrails current with your team.
Deployment options →Security and trust →With Reign Factory →
Who does what.
| Area | Reign and our engineers | Your team |
|---|---|---|
| The gateway | Installs, runs and upgrades the gateway, 24×7. | Holds the model provider accounts. Supplies the models, tools and APIs the gateway connects to. |
| Policy | Writes and updates policies and guardrails with your team. | Sets the policy and approves models, tools and limits. Names policy owners. |
| Identity and registers | Syncs the gateway with your agent and MCP register. | Connects your identity provider. |
| Incidents and evidence | Handles incidents, urgent access changes and key revocation. | Holds the evidence storage. Chooses what content, if any, is captured. |
One subscription, operated for you.
Buy it on its own, or get it included with every Reign Factory subscription.
We configure one governed traffic path with you first, then add the rest.
Operation, policy changes and reporting in one fee.
Our engineers write and update your policies and guardrails with your team.
Questions buyers ask
Do my developers have to change code?
Very little. Point them at one endpoint: OpenAI-compatible, Anthropic-compatible, or MCP. Your SDKs keep working.
Which models can we use?
Your choice, on your own accounts and keys: OpenAI, Anthropic, Gemini, Bedrock, Azure OpenAI, or your own open-weight endpoint.
What gets recorded?
Caller, model, tokens, cost and guardrail decisions for each model call, and each MCP tool call through the gateway. Content capture is off by default.
Who can reach which model or tool?
Identity is checked on every request. Access grants let each person or agent reach only the models and tools you approve.
How are spend and content controlled?
Request, token and dollar limits. Guardrails block or mask sensitive text, and can hand off to a guardrail service you already run.
Where does it run, and who runs it?
In a single-tenant deployment, on Reign Dedicated or in your own AWS or Azure account. It is highly available, and Reign runs, patches and upgrades it 24×7.
Start with a 30-minute governance briefing.
See Reign Gateway apply your policy to one traffic path.