---
schema_version: "newruntime-agent-readable-v0.2"
type: "post"
stable_id: "post:langchain-deep-agents-lean-harness"
slug: "langchain-deep-agents-lean-harness"
title: "LangChain Deep Agents Shrink the Harness Instead of Adding More Prompt"
description: "Deep Agents v0.7.0b2 cuts default-agent input tokens by 65% and tool-description tokens by 43%, turning harness efficiency into a first-class agent metric."
retrieval_nugget: "Deep Agents v0.7.0b2 cuts default-agent input tokens by 65% and tool-description tokens by 43%, turning harness efficiency into a first-class agent metric. LangChain's Deep Agents v0.7.0b2 release is a useful counter-signal to the usual agent pattern: add more instructions, add more tools, add more planning text, then hope a stronger model absorbs the mess."
status: "published"
published_at: "2026-07-27"
updated_at: "2026-07-27"
record_date: "2026-07-27"
date_kind: "published_at"
topics: ["agents","agent-harnesses","evals","developer-tools"]
source_urls: ["https://docs.langchain.com/oss/python/releases/changelog","https://x.com/LangChain/status/2081719358376616128"]
visuals: [{"id":"langchain-deep-agents-lean-harness","kind":"editorial-diagram","role":"hero","src":"https://newruntime.com/images/posts/langchain-deep-agents-lean-harness.webp","alt":"Hand-drawn split diagram showing an overloaded Deep Agents harness trimmed into a lean core with 65 percent and 43 percent token reductions guarded by evals.","caption":"Deep Agents v0.7.0b2 makes harness efficiency visible: smaller prompt and tool surfaces, protected by evaluation.","credit":"New Runtime synthesis from LangChain Deep Agents changelog","source_url":"https://docs.langchain.com/oss/python/releases/changelog","generated_with":"newruntime-style-locked-svg-v1","width":1600,"height":900,"legend":[{"label":"Trim","description":"The release removes duplicated tool prose and starts the authored base prompt empty."},{"label":"65% / 43%","description":"Default-agent input tokens and tool-description tokens both shrink materially."},{"label":"Eval gate","description":"The useful claim is cost and context reduction without quality regression."}]}]
routes: {"html":"https://newruntime.com/posts/langchain-deep-agents-lean-harness/","markdown":"https://newruntime.com/posts/langchain-deep-agents-lean-harness.md","json":"https://newruntime.com/posts/langchain-deep-agents-lean-harness.json"}
source_format: "markdown"
---

# LangChain Deep Agents Shrink the Harness Instead of Adding More Prompt

## Retrieval answer

Deep Agents v0.7.0b2 cuts default-agent input tokens by 65% and tool-description tokens by 43%, turning harness efficiency into a first-class agent metric. LangChain's Deep Agents v0.7.0b2 release is a useful counter-signal to the usual agent pattern: add more instructions, add more tools, add more planning text, then hope a stronger model absorbs the mess.

LangChain's Deep Agents v0.7.0b2 release is a useful counter-signal to the usual agent pattern: add more instructions, add more tools, add more planning text, then hope a stronger model absorbs the mess.

The changelog says the default harness is now leaner. On a default-agent turn, input tokens drop 65% compared with v0.6.12, from 5,395 to 1,895. Tool-description tokens drop 43%, from 4,005 to 2,302. LangChain says this was validated against a revamped evaluation suite with no quality regression.

## What changed

The release trims duplicated tool-usage prose, starts the authored base prompt empty, and makes planning todos opt-in. The filesystem tools also become more practical: better delete/write behavior, tool allowlists, paginated reads, bounded grep/glob output, and a 1,000-match grep cap.

There is also more prompt-caching support and an NVIDIA Nemotron 3 Ultra harness profile.

## New Runtime Read

This is a harness story, not a model story. The agent system becomes better by reducing unnecessary prompt mass and making tool surfaces more explicit.

That matters because token volume is not only a cost line. It is also latency, context pressure, tool confusion, cache behavior, and verification burden. A smaller harness that keeps quality flat is real product work.

The broader pattern: useful agent frameworks will compete on evaluation-backed harness design, not on prompt theater. If a framework cannot show what got smaller, what stayed compatible, and what the eval gate protected, the release is harder to trust.