One underappreciated dimension of building agent platforms: the agent harness — the CLI or framework that wraps the model — matters as much as the model itself. Engineers report meaningfully different results from the same model run through different harnesses. The harness is not a thin wrapper. It is an opinionated layer that shapes agent behavior in ways that matter.
This fragmentation is playing out at every large company. Adoption is split — some engineers use Cursor, some Copilot, some Claude Code, some the open-source CLIs. The Shopify Slack agent was designed from the start to support multiple CLI backends, with code paths for both Pi and Claude even though it is currently only running one. Swapping is a configuration change, not a rewrite. That is the kind of decision that separates systems built by experienced engineers from systems built during a hackathon.
Companies building internal tools are typically locked to one framework because that is what the team that built it knows. A real product has to be agent-agnostic the way good deployment infrastructure is cloud-agnostic. The harness matters. The model matters. Both are moving targets.
Bring-your-own-keys and CLI-agnostic design are table stakes for the mid-market enterprise buyer who refuses to be locked in. The buyer will not commit to a platform that bets on a single harness winning, because no rational buyer believes a single harness will win.
I've argued separately that the convergent agent stack signals a productizable layer. The harness is one of the moving parts inside that stack — and the platform that abstracts the harness, instead of fighting for it, is the one that survives the next twelve months of churn.
Sources
Related Essays
The Framework Cannot Be the Product
Frameworks get commoditized, forked, or absorbed by model providers. The durable revenue layer is the indispensable infrastructure piece that gets harder as the ecosystem matures.
The Agent Harness Problem
Enterprise agents need layered interfaces, real software skills, and flexible platforms. The harness around the model matters more than the model.
Agents Are Software, and Software Needs a Factory
People talk about agent harnesses as if the harness is the interesting part. It is not. The interesting part is the factory — sandboxing, orchestration, persistence, model translation.
Key takeaways
- The agent harness — the CLI or framework that wraps the model — matters as much as the model itself. Different harnesses produce meaningfully different results from the same model.
- Adoption inside large companies is split across Cursor, Copilot, Claude Code, and open-source CLIs. A real platform has to support multiple harnesses by design, not as an afterthought.
- Bring-your-own-keys and CLI-agnostic design are table stakes for the mid-market enterprise buyer who refuses to be locked into a single vendor's harness.
FAQ
What is an agent harness?
The harness is the CLI, framework, or runtime that wraps the model and shapes how it interacts with tools, files, and the user. Claude Code, Cursor, Copilot, and Codex are different harnesses. Same model can behave very differently in each.
Why does harness-agnostic design matter?
Engineers inside the same company use different harnesses, and the best harness for a given task changes monthly. Locking the platform to one harness creates the same problem as locking to one cloud — you are a hostage to the next price hike or feature regression.