Agentic Harnesses Market Landscape, July 2025 to July 2026

Agentic harnesses from July 2025 to July 21, 2026: competitive landscape, market share by use case, model integration, orchestration patterns, and differences between coding harnesses and broader workflow agents, including Claude Code, OpenAI Codex, Cursor, GitHub Copilot, Google Jules, Replit Agent, Devin, Windsurf, OpenCode, Aider, OpenHands, LangGraph, CrewAI, AutoGen, and n8n.

GPT-5.6-sol
financial
frontier
academic
vc
blogs
tech

Synthesised 2026-07-21

Narrative

VC and analyst coverage depicts agentic harnesses as two overlapping but materially different markets. Coding agents are the clearest early product category because repositories offer version control, tests, issue trackers and pull-request review as practical feedback loops. Gartner's September 2025 Magic Quadrant separated AI code assistants as a distinct buyer category, while CB Insights' October market map found 90-plus AI software-development vendors across eight markets and reported that three quarters offered semi-autonomous or autonomous features.

The competitive boundary is increasingly between vertically integrated model-plus-agent products and model-agnostic orchestration layers. Gartner's May 2026 assessment says frontier-model providers have moved directly into the application layer, while its market guide frames the trade-off as integrated performance and automation versus flexibility and workflow integration. Sequoia's Reflection and OpenAI Codex coverage make the integrated case: domain-tuned models, controlled execution environments and repository-specific feedback seek to turn tasks into mergeable pull requests. In contrast, CB Insights and a16z identify orchestration, observability, security, tool integration and durable execution as the expanding infrastructure opportunity around agents.

Public evidence does not support clean product-level market shares for Claude Code, Codex, Cursor, Copilot, Jules, Replit Agent, Devin, Windsurf, OpenCode, Aider or OpenHands. The available measures are uneven proxies: CB Insights reports Anysphere, Replit and Lovable passing $100 million ARR in its October 2025 mapping, and says Google held 57% of hyperscaler software-development partnerships in April 2026, versus 25% for Microsoft and 19% for Amazon. Those figures measure revenue thresholds and partnership portfolios, not active seats, engineering output or durable customer preference. They do, however, support the narrower conclusion that software development led commercial agent activity, while customer service led broader enterprise adoption in CB Insights' Q4 2025 survey.

The broader workflow-agent market remains less mature than coding-agent publicity implies. McKinsey reported in November 2025 that 23% of surveyed organisations were scaling an agentic system somewhere, but no individual function exceeded 10% scaling; IT and knowledge management led reported use. Its 2026 software-delivery work argues that leading teams are redesigning around overnight agent execution, structured requirements, test gates and human review, rather than simply buying a chat interface. Forrester and Gartner supply an important corrective: autonomy varies sharply between products, and multi-agent systems still fall short of many supplier claims. The durable harness requirements are therefore context control, permissions, traces, evaluation, recovery and cost measurement, not merely a planner delegating prompts to several workers.

Sources

ID	Title	Outlet	Date	Significance
v1	Reflection AI Launches Asimov: Breakthrough Agent for Code Comprehension	Sequoia Capital	2025-07	Sequoia's July 2025 investment thesis distinguishes code comprehension from code generation and describes Asimov's retriever-and-combiner multi-agent architecture, permissioned memory and use of third-party models.
v2	The AI agent tech stack	CB Insights	2025-08	CB Insights maps 135-plus startups across 17 infrastructure markets and identifies observability, evaluation, security, marketplaces and cost management as the layers required as agents become autonomous.
v3	Agentic AI Reshapes Software and Services Market Share, Margin and Risk	Gartner	2025-08	Gartner frames agentic systems as shifting value from technology and labour towards outcome delivery, a useful strategic lens for where harness vendors may capture margin.
v4	The Rise of Computer Use and Agentic Coworkers	Andreessen Horowitz	2025-08	a16z separates browser and desktop computer-use agents from scripted RPA, and highlights computer-use models, MCP-style tool access and durable execution as prerequisites for broader workflow automation.
v5	Partnering with Zed: The AI-Powered Code Editor Built from Scratch	Sequoia Capital	2025-08	Sequoia reports Zed's 150,000 monthly active developers and cites survey proxies showing 9% of Rust developers using it, illustrating how developer-tool adoption is measured more often through usage signals than market share.
v6	Gartner Magic Quadrant for AI Code Assistants	Gartner	2025-09	Gartner's September 2025 market definition and vendor set, including Anysphere, Cognition, GitHub and Google Cloud, provides an analyst classification of the enterprise AI code-assistant field.
v7	One year of agentic AI: Six lessons from the people doing the work	McKinsey	2025-09	McKinsey's practitioner report focuses on deployment lessons rather than demos, supporting the view that operating models, workflow redesign and governance constrain real-world agent scaling.
v8	The AI-Native Office Suite – Can AI Do Work For You?	Andreessen Horowitz	2025-09	a16z's task benchmarking of office-oriented agentic products shows that broader workflow agents remain fragmented across generalist and vertical tools, with uneven output quality and execution time.
v9	The AI software development market map	CB Insights	2025-10	CB Insights maps 90-plus private companies across eight software-development markets, reports that three in four offer semi- or fully autonomous features, and argues that planning, implementation, testing and deployment are converging into agentic platforms.
v10	The State of AI: Global Survey 2025	McKinsey	2025-11	McKinsey's survey quantifies the gap between experimentation and scale: 23% of respondents reported scaling an agentic system somewhere, 39% were experimenting, and no function exceeded 10% reported scaling.
v11	Unlocking the value of AI in software development	McKinsey	2025-11	McKinsey's survey of nearly 300 public companies identifies 16% to 30% improvements in productivity, customer experience and time to market among high performers, and describes coding tools evolving into long-running multi-file agents.
v12	The AI agent market map	CB Insights	2025-11	CB Insights maps 400-plus private agent companies across 16 categories, finds software development among the leading revenue areas, and warns that reasoning-model inference costs create pricing pressure for coding-agent vendors.
v13	The Reality of AI Multiagent Systems: Navigating the Hype	Gartner	2025-11	Gartner's November 2025 research explicitly concludes that multi-agent development platforms fall materially short of vendor promises and focuses attention on five implementation pitfalls.
v14	2026: This is AGI	Sequoia Capital	2026-01	Sequoia's January 2026 thesis identifies long-horizon coding agents as a capability threshold and argues that reinforcement learning and agent harnesses are complementary paths, though its AGI framing is investor opinion rather than independent measurement.
v15	Early-Stage Trends Report: Agentic Security, AI Scientists, and more	CB Insights	2026-02	CB Insights finds its tracked cost, security and reliability tooling cohort scored 666 on average versus 588 for coding agents and copilots, indicating investor attention may be shifting towards managing agent-generated code.
v16	5 AI agent predictions for 2026	CB Insights	2026-02	CB Insights uses Q4 2025 enterprise-survey and market data to identify customer service as the leading adoption area and observability, evaluation and continuous red teaming as emerging production priorities.
v17	Hyperscalers are carving up the AI agent market - here is who’s winning where	CB Insights	2026-04	CB Insights reports that Google held 57% of software-development partnerships among the three hyperscalers, compared with Microsoft at 25% and Amazon at 19%, a useful but limited proxy for strategic position rather than user share.
v18	Standard Intelligence: Training General Intelligence in Pixel Space	Sequoia Capital	2026-04	Sequoia's April 2026 thesis argues that video pre-training for mouse, keyboard and screen actions may become an alternative to text-and-tool-call architectures for general computer-use agents.
v19	AI 100: The most promising artificial intelligence startups of 2026	CB Insights	2026-05	CB Insights documents agents executing multi-step enterprise workflows at scale in security and financial-crime operations, while stressing that such agents require governance beyond traditional employee or service-account models.
v20	Gartner Says the Market for Enterprise AI Coding Agents Is Entering a New Phase of Expansion and Competitive Realignment	Gartner	2026-05	Gartner predicts that by 2027 more than 65% of engineering teams using agentic coding will treat IDEs as optional, and identifies governance, pricing and commercial maturity as decisive enterprise criteria.
v21	Rewiring software delivery for the agentic era	McKinsey	2026-05	McKinsey describes organisations redesigning development around daily cycles of human judgement and overnight agent work, claiming threefold to fivefold productivity improvement and 60% smaller teams at leading adopters.
v22	Agentic AI Enters Its Enterprise Execution Era	Forrester	2026-05	Forrester's May 2026 commentary frames the market shift as an execution problem, signalling that enterprises increasingly assess agents through operational deployment rather than demonstrations.
v23	Cost versus value: Managing agentic AI system performance	McKinsey	2026-07	McKinsey analyses 20 production coding workflows and reports that review and refinement accounted for 59.4% of work, showing why verification, context handling and orchestration materially affect agent economics.