Ramp Built an AI Agent Harness in Six Weeks, and LangChain's CEO Says It's Part of a Bigger Shift
The New Stack, Thursday, October 22nd, 2026
Ramp built its own agent harness, Laminar, for model independence and observability, and LangChain's CEO sees a broader trend.
After testing managed agent APIs, open-source frameworks and third-party platforms, fintech Ramp built Laminar, an in-house agent harness that reached a first version in six weeks and now runs all its agentic products.
Ramp wanted to switch models on cost, speed and performance, and to see why agents made unnecessary tool calls or carried bloated context, so it built its own debugging interface.
Laminar keeps sessions separate from sandboxes, stores conversation and execution state independently, and keeps credentials and orchestration outside the sandbox.
LangChain CEO Harrison Chase praised these choices, noting similar designs in LangChain's Deep Agents.