July 2026 Summaries

Speakeasy has introduced Agent Budgets, a new feature designed to manage AI spending limits for teams and individuals, enforced via the AI Control Plane. This tool allows organizations to set specific spending caps or flag overspending for review, providing warnings as limits approach and adjusting rules dynamically. Budgets are applied holistically across all AI usage within a team, covering various tools and providers, ensuring that spending is tracked per person irrespective of the tools used. The budget rules are customizable based on directory attributes such as department or employee type, and can either flag or block requests when limits are reached. The system offers real-time monitoring of budget status, with a detailed breakdown of spending activities and an events feed that records threshold warnings and breaches. Unlike provider-specific limits, Agent Budgets provide a unified approach to managing expenditures across multiple AI tools and providers, ensuring comprehensive financial oversight. The feature is available to organizations with directory sync enabled, and Speakeasy recommends starting with flag rules and warn thresholds for optimal budget management.

Speakeasy offers a comprehensive solution for secure and transparent authorization in AI environments by integrating with a company's identity provider (IdP) and implementing detailed, role-based access controls. Unlike traditional methods that rely on shared service accounts and server-level allowlists, Speakeasy's approach ensures that each agent session is authenticated through a real user, and access permissions are evaluated based on specific policies for each tool. This setup enhances security and accountability by maintaining a detailed identity profile and tracking role assignments, which automatically adjust with team changes. The system uses a multi-step authorization process, involving permissions defined as grants with scopes and selectors, and evaluates each tool call through an additional layer of checks based on the tool's metadata. Deny rules are prioritized to prevent accidental access, and every authorization decision is documented as an auditable record, enabling clear retrospective analysis of access permissions. Speakeasy's integration with directory sync and identity-stamped telemetry ensures that all sessions are linked to real users, and comprehensive audit logs are maintained, facilitating transparency and control over AI-driven operations.

Speakeasy's skills management system addresses the growing need for organizations to organize and version control their agent skills, which are reusable instruction documents in SKILL.md format. As companies increasingly rely on these skills to codify shared working methods, they face challenges with scattered and duplicated skills across various repositories and personal collections. Speakeasy provides a centralized platform to record, version, and browse skills, ensuring that teams have access to the most up-to-date and effective instructions. This system offers features like immutable versioning, validation without gatekeeping, and usage analytics to track and improve skill performance. By consolidating duplicate skills and promoting effective ones, Speakeasy enhances team productivity and ensures that the best practices become standardized across the organization. The platform's AI-native approach allows for performance measurement and continuous improvement of skills, transforming them from isolated solutions into essential team assets.

Speakeasy has launched AI Cost Control, a tool designed to provide organizations with comprehensive visibility into AI spending by capturing real-time usage data across various AI tools like Claude Code, Cursor, and Codex. This system consolidates spending data, differentiating between enterprise and personal licenses, and normalizes session data against OpenTelemetry GenAI conventions to provide a unified cost view. As AI expenses are projected to surpass cloud spending, companies face challenges in tracking costs due to fragmented reporting and diverse tools. Speakeasy addresses these issues by deploying device agents that record AI usage at the machine level, enabling detailed breakdowns of costs by employee, team, and task. The platform also plans to introduce features like budget enforcement, automatic work categorization, and enhanced client coverage to further aid in managing AI expenses.

Speakeasy has introduced a new feature that provides organizations with a unified view of AI usage, distinguishing between personal and team accounts, which is crucial for managing shadow AI and governance. Previously, developers would use personal accounts for AI tools like Claude Max due to cost advantages, resulting in usage that was invisible to company oversight. This update allows organizations to filter and monitor AI usage across different account types, ensuring that even personal account activity is visible and subject to the same governance as enterprise accounts. The system classifies accounts by resolving session emails against organizational membership and uses identifiers to maintain a clear distinction between personal and enterprise usage. This capability is essential for compliance with frameworks such as ISO 42001 and SOC 2, as it allows organizations to audit AI tool usage and maintain records of AI operations, addressing a significant blind spot in AI governance.