Speakeasy service tracks enterprise-wide AI agent spending
Speakeasy Development Inc. today introduced an artificial intelligence cost-management service intended to give companies a consolidated view of spending across coding agents and other AI tools.
The startup’s AI Cost Control service collects usage data from tools including Anthropic PBC’s Claude Code, Anysphere Inc.’s Cursor, Claude Cowork and OpenAI LLC’s Codex. It records the tokens consumed, model used, cache activity and estimated cost of individual interactions, then aggregates the information by employee, team, tool and account type.
The service is part of Speakeasy’s AI Control Plane, which sits between employees and the AI tools they use. Speakeasy said it addresses a growing financial management problem as companies adopt multiple AI services, each with its own usage measurements, billing practices and administrative interface.
“The problem enterprises have is that the model provider’s numbers live in silos,” said Chief Executive Sagar Batchu. “Anthropic can tell you what your Anthropic key spent, but it can’t provide a unified view across competing models and tools.”
That fragmentation is complicated by employees’ use of personal subscriptions for company work. Enterprise reports generally capture only centrally managed accounts, potentially leaving organizations unaware of activity conducted through individually purchased plans.
Speakeasy deploys a software agent on employees’ work devices to collect AI session telemetry. It associates the activity with an employee’s work identity even when the AI service is accessed through a personal account. The company said it does not connect to employees’ personal accounts or collect activity conducted on their personal devices.
“We observe real-time AI activity on the enrolled work machine, not the personal account itself,” Batchu said. “We never touch the account’s history from their personal devices.”
Protected data
While collecting session data, including transcripts, could raise privacy and security questions, Batchu described it as a form of workforce observability and said Speakeasy has agreements governing how customer data is stored and used. The company also offers dedicated data residency to selected enterprises and deployments in customers’ virtual private clouds. Data collection follows the OpenTelemetry Semantic Conventions for generative AI, which provide a common format for such information as token counts, models, costs and tool-call metadata. Speakeasy normalizes those records to produce a consistent view across providers.
AI Cost Control currently provides per-turn cost estimates, cross-agent consolidation and breakdowns by factors such as employee, team, model, session and source. Customers can drill into individual sessions to identify the tools, skills and Model Context Protocol servers responsible for consumption.
Speakeasy differentiates the service from functionally similar AI gateways and observability products by focusing on employee-operated AI clients. Batchu said gateways can track requests routed through them, but much of the AI use in an organization occurs through desktop applications and subscription services that bypass those systems.
“Observability platforms instrument your own application’s model calls,” he said. “They answer ‘What does my product spend?’ not ‘What does my workforce spend?’”
The company plans to add enforceable budgets for teams and individual employees. That will give customers notifications and require approval for continued high usage cases or restrict activity routed through Speakeasy’s gateway when a threshold is reached.
Other planned features include automated categorization of spending by tasks such as code review, research and ticket triage, along with separate measurement of tokens consumed when an agent loads system prompts, skills and MCP servers. Speakeasy also intends to reconcile flat-rate subscription costs with metered application programming interface charges.
The system does not currently capture work performed on unmanaged personal devices or AI features embedded in software-as-a-service applications. Speakeasy said coverage of embedded AI services is on its roadmap.
Image: SiliconANGLE/Microsoft Designer
Support our mission to keep content open and free by engaging with theCUBE community. Join theCUBE’s Alumni Trust Network, where technology leaders connect, share intelligence and create opportunities.
15M+ viewers of theCUBE videos, powering conversations across AI, cloud, cybersecurity and more** 11.4k+ theCUBE alumni**— Connect with more than 11,400 tech and business leaders shaping the future through a unique trusted-based network.
About SiliconANGLE Media
theCUBE AIand theCUBE SuperStudios — with flagship locations in Silicon Valley and the New York Stock Exchange — SiliconANGLE Media operates at the intersection of media, technology and AI.
Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands that reach 15+ million elite tech professionals. Our new proprietary theCUBE AI Video Cloud is breaking ground in audience interaction, leveraging theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of industry conversations.