Your dev teams are already using a coding agent you haven't approved

Open Code's 20x growth to 13 million monthly actives and 7 trillion daily tokens processed shows model-agnostic coding agents are becoming the norm. Its pay-per-token revenue run rate of $40M proves real willingness to pay for flexibility.

When Anthropic blocked Open Code from using Claude subscriptions, the backlash equated the two products and spurred massive adoption. Later, enterprises begged for security paperwork because hundreds of employees already used it—bottom-up adoption has bypassed procurement entirely.

Global usage data reveals cost-sensitive markets rely on cheaper open-source models like DeepSeek Flash to extend coding agent access. Even US enterprises explore model diversity to avoid lock-in and control spend, enabled by volume discounts on inference.

For engineering leaders, this creates an urgent governance challenge: employees select tools without oversight, forcing a tradeoff between innovation velocity and cost control. Leaders must build guardrails around token spend, security, and model selection while deciding whether to standardize or embrace multi-model flexibility.