Ramp builds internal AI agent harness Laminar in six weeks

What's Happening

• Fintech company Ramp developed Laminar, an internal AI agent harness, over six weeks to power its agentic products, according to The New Stack.
• The platform consolidates agent execution, sandboxing, observability, and asynchronous workflows into a single system detailed in a Tuesday engineering post.
• Ramp engineers built the system to avoid vendor lock-in and maintain the flexibility to switch model providers based on cost and performance, the company noted.

Why It Matters

Enterprises face a build-versus-buy decision for AI agent infrastructure. Ramp's choice to develop Laminar highlights the limitations of current third-party platforms in providing adequate observability and model flexibility. Custom harnesses give product teams direct control over debugging and execution environments.