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

The commercial coding-harness market concentrated rapidly around cloud agents, IDE-native assistants and terminal tools. Bloomberg reported that Cursor had surpassed $500m in annualised revenue by June 2025, while its March 2026 coverage put the company at more than one million daily users and 50,000 business customers. OpenAI, Anthropic, Google and GitHub each paired frontier models with managed execution environments, repository context and enterprise administration, reducing the advantage that standalone code editors gained from model access alone.

The sources distinguish coding harnesses from general workflow agents by their execution surface and governance requirements. Codex, Jules and GitHub Copilot coding agent run delegated repository work in isolated cloud environments and return diffs or pull requests for review; Claude Code centres on terminal access, permissions and sandboxing. By contrast, LangGraph and n8n position themselves as orchestration layers for multi-step, cross-application processes, where observability, state, evaluations and tool connectivity matter more than repository-aware editing.

Public adoption measures remain uneven and frequently vendor-reported. JetBrains found in January 2026 that 90% of developers regularly used at least one AI tool at work, while GitHub Copilot remained the most widely known and adopted individual coding product in its survey. A January 2026 research census estimated coding-agent adoption across 129,134 GitHub projects at 15.85% to 22.60%, and a later 180-million-repository study found different adoption signatures for cloud pull-request agents versus local or commit-oriented tools, underscoring why revenue, stars, user counts and commit traces should not be treated as interchangeable market share.

Investment followed the products that could package autonomy with enterprise procurement. Cognition said in May 2026 that it had raised more than $1bn at a $26bn valuation and had $492m in revenue run rate, while n8n said its October 2025 Series C valued it at $2.5bn. These figures show investor demand for both autonomous software engineering and business-process orchestration, but they are company statements rather than audited measures of task success, retention, or displacement of human work.

Sources

ID	Title	Outlet	Date	Significance
f1	OpenAI Takes on Google, Anthropic With New AI Agent for Coders	Bloomberg	2025-05	Bloomberg records the May 2025 launch of Codex as a research-preview cloud agent for writing features, fixing bugs and running tests, placing OpenAI in a crowded coding-agent market.
f2	AI Coding Assistant Cursor Draws a Million Users Without Even Trying	Bloomberg	2025-04	Bloomberg documents Cursor's editor-first product, support for OpenAI and Anthropic models, and word-of-mouth penetration across companies including OpenAI, Instacart, Spotify and Uber.
f3	Anysphere, Hailed as Fastest Growing Startup Ever, Raises $900 Million	Bloomberg	2025-06	Bloomberg reports Cursor's claimed $500m annualised revenue, $900m financing and $9.9bn valuation, a key financial indicator of demand for AI coding tools.
f4	Google Debuts Gemini AI Coding Tool in Bid to Entice Developers	Bloomberg	2025-06	Bloomberg positions Gemini CLI as Google's response to Codex and Claude Code, showing frontier labs moving coding agents into the developer terminal.
f5	AI Startup Cognition to Buy Windsurf After Google Licensing Deal	Bloomberg	2025-07	Bloomberg details the July 2025 split transaction in which Google licensed Windsurf talent and technology for $2.4bn and Cognition bought the remainder, evidence of strategic competition for coding-agent assets.
f6	AI Coding Agents Like Claude Code Are Fueling a Productivity Panic in Tech	Bloomberg	2026-02	Bloomberg Businessweek captures the organisational pressure and productivity claims around increasingly autonomous coding agents, rather than treating adoption as a purely technical shift.
f7	AI Coding Startup Cursor Plans New Model to Rival Anthropic, OpenAI	Bloomberg	2026-03	Bloomberg reports Cursor's planned Composer 2 model, more than one million daily users and 50,000 businesses, illustrating both its move from model aggregation towards proprietary models and its enterprise scale.
f8	AI Coding Startup Cursor in Talks for About $50 Billion Valuation	Bloomberg	2026-03	Bloomberg provides a later capital-markets measure of investor expectations for the coding-agent category, with Cursor reportedly seeking a valuation near $50bn.
f9	Introducing Codex	OpenAI	2025-05	OpenAI's primary launch document defines Codex as a cloud software-engineering agent that runs parallel tasks in separate repository-preloaded sandboxes and proposes pull requests for review.
f10	Codex is now generally available	OpenAI	2025-10	OpenAI describes the October 2025 enterprise product surface, including Slack, SDK, administration, monitoring and analytics, plus usage and customer claims that show the push from individual tool to managed engineering platform.
f11	Introducing the Codex app	OpenAI	2026-02	OpenAI's product note describes a transition from single-agent edits to coordinated multi-agent supervision, worktrees and persistent cloud automations.
f12	Build with Jules, your asynchronous coding agent	Google	2025-05	Google explains that Jules clones repositories into a secure cloud VM, uses Gemini 2.5 Pro and returns plans and diffs, making its asynchronous cloud-agent architecture comparable to Codex and Copilot coding agent.
f13	Gemini CLI: your open-source AI agent	Google	2025-06	Google frames Gemini CLI as an open-source terminal agent integrated with Gemini Code Assist and available through personal, AI Studio and Vertex AI access routes.
f14	Getting started with GitHub Copilot coding agent is now easier	GitHub	2025-06	GitHub explains its delegated coding-agent workflow, which executes in a GitHub Actions cloud environment and returns work for review under repository and enterprise controls.
f15	Claude and Codex now available for Copilot Business & Pro users	GitHub	2026-02	GitHub shows the strategic shift towards a multi-model agent control plane, offering Claude, Codex and Copilot inside shared repository context, policy and audit controls.
f16	Beyond permission prompts: making Claude Code more secure and autonomous	Anthropic	2025-10	Anthropic's engineering account makes sandboxing, filesystem and network boundaries central to Claude Code's autonomy and reports an internal 84% reduction in permission prompts.
f17	Claude Code and new admin controls for business plans	Anthropic	2025-08	Anthropic documents how premium seats, compliance APIs, usage analytics, tool permissions, file restrictions and MCP configuration turn Claude Code into an enterprise-managed product.
f18	Which AI Coding Tools Do Developers Actually Use at Work?	JetBrains	2026-04	JetBrains supplies a cross-tool adoption survey, reporting that 90% of developers regularly used AI tools at work by January 2026 and that GitHub Copilot remained the best-known and most-used individual product.
f19	Agentic Much? Adoption of Coding Agents on GitHub	arXiv	2026-01	This research paper estimates coding-agent adoption at 15.85% to 22.60% across 129,134 GitHub projects and distinguishes high-autonomy agents from earlier completion tools.
f20	Detecting AI Coding Agents in Open Source: A Validated Multi-Method Census of 180 Million Repositories	arXiv	2026-06	The large repository census provides a useful corrective to vendor usage metrics, distinguishing cloud pull-request agents from local or commit-oriented agents and identifying different work patterns.
f21	State of Agent Engineering	LangChain	2026-06	LangChain's survey of more than 1,300 practitioners reports 57.3% production deployment, widespread multi-model use, and strong emphasis on observability and evaluation for broader agent systems.
f22	LangGraph Platform is now Generally Available: Deploy & manage long-running, stateful Agents	LangChain	2025-05	LangChain's platform announcement identifies the broader workflow-harness features absent from simple coding assistants: persistent state, asynchronous execution, human-in-the-loop operation and state inspection.
f23	n8n raises $180m to get AI closer to value with orchestration	n8n	2025-10	n8n reports a $180m Series C, $2.5bn valuation, sixfold user growth and tenfold revenue growth, providing a financial signal for model-agnostic workflow orchestration rather than code generation alone.
f24	More Devins in More Places	Cognition	2026-05	Cognition claims more than $1bn raised at a $26bn valuation, $492m revenue run rate and tenfold enterprise usage growth, making it a central but vendor-reported measure of autonomous software-engineering demand.
f25	OpenHands Enterprise	OpenHands	2026	OpenHands documents the open and self-hosted alternative: isolated container execution, bring-your-own model providers, private-cloud deployment and enterprise role controls.