UC Berkeley's Agency Cuts Agent Harness Costs by Up to 35%
The practical shift here is that cost optimisation moves from model selection to infrastructure — a middleware layer can be reused across Claude Code, Codex and other harnesses. Engineering leads running production agents should audit trace logs for repeated context reads and unnecessary tool calls before upgrading to a costlier model. Teams in cost-sensitive markets, including ASEAN startups paying USD-denominated API bills, gain a lever that compounds across every harness they deploy.
UC Berkeley researchers found that changing an agent's harness — the prompts, tools and orchestration around a model — barely affects accuracy but changes cost by up to 2.6x on the same model and task. The takeaway: to lower agent spending, change the harness and keep the model, rather than paying for a pricier or larger model.
The team analysed agent traces and identified five ways to cut cost, each working by reducing how much context an LLM call reads or how many calls a task requires. They built Agency, a middleware layer that sits between existing harnesses such as Claude Code and Codex and the model, so an optimisation written once works across different harnesses. Their five proof-of-concept techniques cut cost by up to 35% with marginal accuracy loss.
The research matters for teams whose agent bills keep growing: auditing repeated context reads and redundant tool calls can yield savings without switching models or sacrificing output quality.
Great practical paper on optimizing harness costs.
For lower agent costs, you want to change the harness and keep the model.
UC Berkeley researchers found that harness changes such as prompts and tools barely move accuracy, but they change cost by up to 2.6x on the same model and task.
They analyzed agent traces and found five ways to cut cost. Each one reduces how much context an LLM call reads or how many calls a task needs.
Then they built Agency, a layer that sits between existing harnesses like Claude Code and Codex and the model. An optimization written once works across harnesses.
Five proof-of-concept techniques cut cost by up to 35% with marginal accuracy loss.
If your agent bill keeps growing, look at repeated context reads and extra tool calls before switching models.
https://t.co/aaLRm7tOLw
Chat with Paper: https://t.co/FQgxm3IHgu
This report is categorized under AITOP's "Papers" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Agency, UC Berkeley, agents, coding assistants.
Source attribution: curated from DAIR.AI (@dair_ai).
