C1.ai Launches a Gateway to Control Where Company AI Requests Go

Companies using C1.ai’s new gateway can put rules between an employee or AI agent and the model receiving their request. C1.ai launched C1 LLM Gateway on October 1, 2026, with immediate availability. The company says it combines controls over where data goes with records of who called a model and what that call cost—making access and spending part of the same decision.

C1 frames the launch as a response to enterprise AI traffic growing faster than the controls around it. In its account, requests can follow default provider connections rather than explicit rules about which companies may receive particular data. Costs can also reach the finance team as one provider invoice, without a clear breakdown by the person, agent or application responsible.

The company identifies two consequences: sensitive information may reach an unintended provider, and overspending may become visible only after the bill arrives. It also argues that default routing can deepen dependence on a single model company. Those are C1’s reasons for introducing a policy layer, not measured findings about customers using the new product.

The gateway is the route through which users, agents and applications call AI models. C1 says its routing policy determines which models each identity is allowed to reach. The security team owns that policy, so the permitted destination is an organizational rule rather than simply a default model connection.

C1 says sensitive work routes to infrastructure the customer controls. Organizations can also prevent selected data from reaching selected providers altogether. Other work can be directed according to speed or cost. The design therefore separates permission from optimization: some destinations are ruled out by policy, while other requests can be routed for a faster or cheaper response.

Policy should decide which models see which data, and you should know what every one of those calls costs and who made it.

Alex Bovee, CEO and co-founder of C1.ai

Once a request passes through the gateway, C1 says it records the caller, the model that answered and the resulting cost. That connects spending to a user, agent or application rather than leaving it only in a provider-level total. The product also includes alerts intended to arrive before a spending overage.

The spending controls extend beyond a dashboard. C1 describes model budgets as governed like access rights, with a request-and-approval process when more capacity is needed. Application owners approve increases, while model access can carry an expiration or be withdrawn. The announced controls cover three points in that process:

• When a budget runs out, users can request an increase for an application owner to approve.
• An authorization can be time-limited rather than left open indefinitely.
• Revocation takes effect immediately, C1 says: the next model call fails.

C1 LLM Gateway is part of C1 Run, which the company describes as its runtime layer for secure agents. It is the fourth and final release of C1 Launch Week. C1 places it alongside C1 MCP Gateway, with the two products applying identity and policy to model calls and tool calls, respectively.

The launch announcement does not disclose pricing or explain how the gateway identifies sensitive requests. That leaves buyers without two important details: what the control layer costs and how its data-handling rules are applied. The release describes routing, attribution and access controls, but supplies no measured savings or performance results to establish their effect in deployed workloads.