Cribl has unveiled StreamAI, an innovative AI capability designed as an enterprise AI gateway integrated with Cribl’s open data platform. This enhancement allows organizations to efficiently manage AI data traffic, token usage, costs, and associated risks through an advanced model router.
StreamAI, developed after evaluating the performance of various AI models on telemetry data, introduces a model router that optimizes AI inference. It smartly directs AI prompts and workloads to the appropriate model based on cost and requirements, enhancing visibility and control over AI operations. In support of this initiative, Cribl offers free inference for clients utilizing automatic data routing to benchmarked AI models, preventing unexpected token expenses.
Cost Control in AI Inference
As companies increasingly incorporate AI into their daily functions, predicting and managing AI inference expenditure has become challenging.
Infrastructure costs can unexpectedly surge, particularly when AI agents manage complex tasks independently. StreamAI addresses these concerns by ensuring high-performance, real-time inference on both proprietary and open-source models, simultaneously enforcing budget constraints and spending limits. Clint Sharp, co-founder and CEO of Cribl, remarked, “AI is not one-size-fits-all. The right model depends on the job, the context, and the economics of the request.”
Adaptive Model Routing
StreamAI enhances each AI interaction's security by protecting sensitive data
StreamAI’s model router is informed by Cribl’s SecIT Bench research, which assessed 20 AI models through 30 real-world IT and security incidents. The research highlighted a 17% difference in diagnostic accuracy against a 20-fold variation in investigation costs, emphasizing that the most costly model isn't always the optimal choice. StreamAI ensures uninterrupted usage by switching to alternative models when an application reaches its cost threshold, with circuit breakers that manage token consumption to avoid budget excesses.
Security and compliance are core to StreamAI, incorporating controls directly into its routing layer. It extends beyond Cribl-Privacy, the telemetry-focused model behind Cribl Guard, to safeguard high-volume production telemetry data. StreamAI enhances each AI interaction's security by protecting sensitive data, blocking potentially malicious requests, and regulating model access. It also features bidirectional redaction of sensitive information in prompts and responses and documents all model interactions as standardized, audit-ready telemetry, offering organizations comprehensive oversight and governance capabilities.
StreamAI is crafted to assist organizations in adopting AI without committing to a singular model, provider, or platform. While StreamAI’s official release is forthcoming, interested customers can contact their Cribl account team for early access.
Cribl, the AI Platform for Telemetry, announces StreamAI, the latest addition to its AI capabilities: an enterprise AI gateway with a model router built into Cribl’s open data platform. Cribl brings its proven ability in managing and routing data at a massive scale to help organizations control runaway AI traffic, token usage, costs, and risk.
After evaluating how various AI models performed on telemetry data, Cribl built a model router to enable more efficient AI inference. StreamAI routes AI prompts and workloads to the right model at the right price while gaining visibility and control over AI. Cribl is offering free inference for customers using automatic data routing to benchmarked AI models, eliminating surprise token bills.
High-performance inference
As more enterprises ramp up AI use in daily operations, spending on AI inference is becoming harder to predict and control. Infrastructure costs can spike with little warning, particularly when agents are coordinating multi-step tasks with minimal human intervention. StreamAI delivers real-time cost control by providing reliable, high-performance inference across proprietary and open-source models while enforcing budgets and spending cutoffs.
“AI is not one-size-fits-all. The right model depends on the job, the context, and the economics of the request,” said Clint Sharp, co-founder and CEO of Cribl. “Customers shouldn’t have to send every prompt to the most expensive model or build their AI strategy around a single provider. StreamAI gives them an intelligent control plane that routes each request to the model best suited to the work, helping reduce token costs while preserving the choice and flexibility to use the models they want.”
Most expensive model
StreamAI’s model router builds on Cribl’s SecIT Bench research, which evaluated 20 AI models across 30 real-world IT and security investigations. It gave AI models the same messy incident data, such as a security breach, service outage, or performance problem.
Initial results revealed a striking insight: a 17% spread in diagnostic accuracy compared with a 20x range in investigation spend, proving that the most expensive model is not automatically the best fit for every workflow. When an AI-powered enterprise application reaches a cost limit, StreamAI automatically falls back to an alternative model to keep usage running without disruption. Built-in circuit breakers give IT teams control over token consumption, helping prevent applications from blowing through quotas and costs.
Shielding proprietary information
To ensure enterprise readiness, StreamAI incorporates critical security and compliance controls directly into the routing layer. It builds on Cribl-Privacy, the custom telemetry-focused model behind Cribl Guard’s background detection, which is designed for the high-volume, semi-structured data found in production telemetry.
StreamAI applies security controls throughout every AI interaction, helping protect sensitive data, block malicious requests, control what models can access, and reduce the risk of unsafe responses or actions. It also provides bidirectional sensitive data redaction, shielding proprietary information, credentials, and sensitive data in both outgoing prompts and incoming responses. Additionally, it automatically records every model call and routing decision as normalised, audit-ready telemetry, giving organizations complete visibility and evidence for governance across all routed AI traffic.
StreamAI is designed to help organizations adopt AI without locking their observability strategy to one model, provider, or closed platform. StreamAI will be available soon. Customers interested in early access can contact their Cribl account team.
