Google Cloud to start charging for Vertex AI agent development tools in 2026

Google Cloud will begin billing for Vertex AI Agent Engine Sessions, Memory Bank, and Code Execution on February 11, 2026. The start date was pushed back from January 28 to give users more preparation time. These components were previously free.

Agent Engine Runtime has offered a free tier since March 2025 (first 50 vCPU-hours and 100 GiB-hours monthly); usage beyond that is billed. New rates are:

• Sessions: $0.25 per 1,000 events stored.
• Memory Bank: $0.25 per 1,000 memories stored per month; $0.50 per 1,000 retrieved (first 1,000 retrievals free each month).
• Code Execution: $0.0864 per vCPU-hour and $0.009 per GiB-hour.

Billing rounds to the nearest second and excludes idle time. The delay provides two additional weeks to audit resources, delete unused state, and update budget forecasts. Note that Runtime pricing was lowered in December 2025.

Why It Matters

• Audit Agent Engine resources now: delete unused Sessions and Memory Bank entries to avoid charges when billing starts on Feb 11.
• Model monthly costs using the new per-event, per-memory, and per-hour rates to update forecasts and budgets.
• Use free tiers and the first 1,000 free Memory Bank retrievals during prototyping to limit spend while adjusting workloads.
• Leverage per-second billing for short sandboxed code runs; consolidate longer jobs to control vCPU- and GiB-hour costs.