Backplanes provides a comprehensive observability platform for AI coding agents, specifically designed to help engineering teams understand and manage the actions performed by tools like Claude Code and Codex. By capturing every agent session, Backplanes generates detailed reports that summarize what was built, which files were modified, and whether any security boundaries were crossed. It acts as an essential layer of visibility for teams increasingly relying on automated agents to execute tasks. The platform aims to bridge the gap between rapid agent-driven development and the necessity for human oversight, governance, and cost control.

The functionality centers on the automated analysis of agent transcripts once a session concludes. Instead of manual review, Backplanes provides a concise verdict for each session, highlighting key findings, time allocation, and potential security risks. This allows developers to quickly assess whether their agents are working correctly and enables managers to track engineering output, spend, and team velocity without needing to read through raw log files or pull requests.

Some of the key features are:

• Session Reports: Automated analysis of every agent run providing a verdict, time breakdown, and specific findings such as credentials written to disk or scope drift.
• Org-wide Visibility: Consolidated reporting that rolls up individual session data into actionable views for Security, Engineering, and Finance stakeholders.
• External Access Inventory: Real-time tracking of all external domains, MCP servers, tools, and skills reached by agents, with capabilities to sanction or block unknown dependencies.
• Cost Attribution: Granular visibility into token spend, attributed by engineer, repository, and model, allowing for accurate financial forecasting and reconciliation with invoices.
• Local Data Redaction: Advanced privacy-preserving measures that strip secrets and PII on the user's machine before data ever leaves the local environment, ensuring sensitive information is never transmitted.
• Zero Retention Policy: Contractual zero-data retention agreements with LLM providers to ensure that any session content analyzed by models for insights is never stored by third parties.

Operationally, the tool is installed via a simple CLI command on macOS, Linux, or WSL 2 environments. Once installed, it automatically captures completed agent sessions, performs local redaction, and processes the telemetry. Users can then view detailed insights through the Backplanes web dashboard, where findings are categorized, triaged by severity, and prioritized based on impact. The process is designed to be seamless, requiring no changes to the agent workflow itself.

Some common use cases include:

• Security Audits: CISOs can use the platform to identify and block unauthorized external domain access or monitor for data egress violations across the entire organization.
• Engineering Oversight: Engineering managers can gain visibility into where team capacity is being spent, which repositories have the most agent activity, and which engineers may require additional support.
• Cost Management: CFOs and financial leads can track AI token spending, identify cost outliers, and reconcile agent-driven expenditures against monthly cloud invoices.
• Developer Review: Engineers can use session reports to quickly verify that their agents followed best practices, performed intended tasks, and didn't introduce accidental regressions or security issues.