Part of the SD Times 100 2026 series. See the full SD Times 100 2026 list for every category and honoree.
No category in this year’s SD Times 100 generated more debate internally than this one, and not because the companies in it are unworthy. It’s because “AI coding tool” stopped being a useful description sometime in the last eighteen months. What started as autocomplete-on-steroids has become something closer to a new layer of the development stack itself: AI that plans features, writes and tests code across entire repositories, reviews pull requests, and in some workflows, ships changes with only light human supervision. We’re calling this year’s theme the Era of Agentic Infrastructure for exactly this reason. The story isn’t “AI is helping developers write code faster.” The story is that AI is becoming a participant in the software development lifecycle with its own permissions, its own failure modes, and its own governance requirements.
For engineering and development leaders, this category is no longer optional reading. Decisions made here — which models, which tools, which guardrails — now shape engineering velocity, code quality, security posture, and hiring strategy all at once.
Why This Category Matters Now
The unit of work is shifting from “lines of code” to “outcomes delegated.” Senior developers increasingly describe their day-to-day not as writing code, but as describing what they want built, reviewing what an AI system produced, and deciding whether to accept, redirect, or escalate. This is a genuine change in the nature of software engineering work, and engineering leaders who haven’t adjusted how they evaluate productivity, code review practices, and onboarding for this reality are already behind.
Model and tool choice has become a strategic decision, not a developer preference. A year ago, picking an AI coding assistant was largely a matter of individual taste. Today, the choice of underlying model and tooling affects security review requirements, IP and code-retention policies, cost at scale, and how well a tool integrates with an organization’s existing CI/CD, version control, and project management systems. This has pulled the decision up from individual contributors to engineering leadership and, in many organizations, to procurement and security teams.
Governance and guardrails are now a first-class engineering concern. As AI tools move from suggesting code to autonomously executing multi-step tasks across a codebase, the question of “what is this system allowed to touch, and how do we know what it did” has become as important as the productivity gains themselves. Feature flagging, audit trails, and staged rollout capabilities that used to be relevant mainly for human-deployed changes are now essential for AI-initiated ones too.
The Different Segments Inside This Category
This category spans a wider range of company types than almost any other in the SD Times 100, and it’s worth separating them clearly.
Foundation model providers. Anthropic, Google, OpenAI and Amazon sit at the base of this stack, building and operating the large language models that power coding assistants, agents, and increasingly, the reasoning layers behind developer tools made by other companies in this same list. Development leaders need a point of view not just on which tools to use, but on which underlying models those tools run on, since model choice affects everything from code quality and hallucination rates to data handling and regional compliance.
AI-native IDEs and editors. Cursor and Windsurf represent the segment that’s grown fastest: full development environments built from the ground up around AI-assisted and increasingly agentic coding, rather than AI bolted onto an existing editor. JetBrains represents the other path, integrating deep AI capability into IDEs that already had a massive, loyal developer base built over two decades.
Embedded coding assistants inside existing platforms. Microsoft GitHub Copilot remains the most widely deployed example here, embedded directly into the version control and collaboration workflow most enterprise developers already use daily. The advantage of this segment is adoption: developers don’t need to change tools, just turn on a capability inside one they already use.
No-code and natural-language application builders. Lovable and Bolt.new, and v0.dev represent a genuinely new segment for this list: tools that let people describe an application in natural language and get a working, deployable product, collapsing the distance between idea and shipped software for prototypes, internal tools, and increasingly, production applications. The three differ in emphasis: Lovable leans toward full-application generation, Bolt.new toward in-browser scaffolding and rapid iteration, and v0.dev toward generating production-ready UI components that plug into an existing codebase. This raises real questions for engineering leaders about where these tools fit relative to professional engineering teams, and when “vibe-coded” software needs to graduate into a properly engineered codebase.
Deployment, runtime, and shipping infrastructure for AI-built software. Vercel occupies an important connective position: as more code (AI-generated or human-written) needs to ship quickly and reliably, the platforms that handle deployment, preview environments, and runtime become part of the AI development story, not separate from it.
Feature management and progressive delivery. LaunchDarkly earns its place here because feature flagging has become essential infrastructure for safely rolling out AI-assisted and AI-generated changes, letting teams ship faster with AI assistance while retaining the ability to instantly roll back if something goes wrong.
Enterprise collaboration and work management with embedded AI. Atlassian represents how AI is being woven into the surrounding fabric of how engineering teams plan, track, and collaborate on work, not just the code itself. Hugging Face has created an open-source model and tools that the machine learning community uses to collaborate on models, datasets, and applications.
Version control, asset management, and enterprise software lifecycle. Perforce and Progress represent the enterprise software lifecycle and digital experience side of this category, where AI capability is being layered into established platforms that already manage source code, large binary assets, or application development at scale.
Agent orchestration and reasoning frameworks. LangChain sits in a distinct and increasingly critical segment: the frameworks and tooling developers use to actually build AI agents and orchestrate multi-step reasoning, rather than tools developers use to write code with AI assistance. As more organizations build their own agentic systems rather than only consuming someone else’s, frameworks like this become foundational infrastructure in their own right.
Multi-agent orchestration at enterprise codebase scale. Block (2026 Addition) represents a newer and distinct problem from single-repository coding assistance: coordinating multiple AI agents working across a large, multi-service codebase at once. Built on Block’s open-source goose agent framework, this segment addresses what happens when AI coding tools that work well in one repository need to operate reliably across an organization’s full surface area of services, without a human manually directing every agent individually.
The most mature organizations are no longer treating AI coding tools as something individual developers opt into on their own. They’re standardizing on a small set of approved tools, often tiered by task: a fast, embedded assistant for everyday code completion and small changes; a more powerful agentic tool for larger, multi-file tasks like feature implementation or refactoring; and increasingly, a separate orchestration layer for building AI agents that are part of the product itself, not just part of the development process.
A pattern worth watching closely: code review practices are being rewritten specifically to account for AI-generated code. Some organizations require a different review checklist for AI-assisted pull requests, with specific attention to whether generated code introduced security issues, license or IP concerns, or subtle logic errors that look plausible but are wrong. Others are investing in AI-powered review tools specifically to keep pace with the volume of code now being produced, since human review throughput hasn’t scaled at the same rate as code generation.
The other significant shift is in how teams think about junior developer onboarding and skill development. If AI tools can produce working code quickly, the differentiating skill for engineers becomes the ability to specify problems clearly, evaluate AI output critically, and debug systems they didn’t personally write line-by-line. Engineering leaders are increasingly building this explicitly into how they train and evaluate junior talent, rather than assuming it develops naturally.
- What happens to our code and data? Model providers and tool vendors differ significantly in data retention, training-on-customer-code policies, and regional hosting options. This is now a procurement and legal question as much as a technical one.
- How agentic is the tool, and what’s the blast radius if it gets something wrong? A tool that can autonomously modify multiple files, run commands, or deploy code needs commensurately strong audit trails, permission scoping, and rollback capability.
- Does it fit the existing toolchain, or require wholesale migration? Embedded assistants inside existing IDEs and platforms have a lower adoption cost than AI-native tools that ask developers to change their daily environment, even if the AI-native tools are more capable.
- How is productivity actually measured, and is the tool’s vendor data trustworthy? Vendor-reported productivity gains should be treated skeptically until validated against an organization’s own before-and-after data, ideally using engineering intelligence tooling rather than self-reported developer sentiment alone.
The 2026 Honorees in Intelligence-Augmented Development
- Anthropic — Foundation model provider powering coding assistants and agentic developer tools.
- Atlassian — Collaboration and work management platform embedding AI across planning and development workflows.
- Microsoft GitHub Copilot — AI coding assistant embedded directly into the world’s largest source control platform.
- JetBrains — AI-enhanced IDEs serving a large, established professional developer base.
- LaunchDarkly — Feature management and progressive delivery platform for safely shipping AI-assisted changes.
- Perforce — Version control and asset management for large-scale, complex enterprise development.
- Progress — Enterprise application development and digital experience platform.
- Vercel — Deployment and runtime platform for shipping web applications quickly and reliably.
- Amazon — Cloud and foundation model provider supporting AI-assisted development at enterprise scale.
- Google — Foundation model and cloud provider powering AI coding and agentic tooling.
- Hugging Face (2026 Addition) — Collaborative platform providing open-source tools and models for building AI applications.
- Cursor (2026 Addition) — AI-native code editor built around agentic, multi-file coding workflows.
- Windsurf (2026 Addition) — AI-native IDE focused on agentic development and codebase-wide reasoning.
- Lovable (2026 Addition) — Natural-language application builder for rapidly turning ideas into working software.
- Bolt.new (2026 Addition) — In-browser, natural-language application builder focused on rapid scaffolding and iteration.
- OpenAI (2026 Addition) — An AI research and deployment company that creates advanced systems like ChatGPT and GPT-4.
- LangChain (2026 Addition) — Framework for building and orchestrating AI agents and multi-step reasoning systems.
- Block (2026 Addition) — Multi-agent orchestration layer for coordinating AI coding agents across large, multi-service codebases.
Frequently Asked Questions
What does “agentic” actually mean in the context of AI coding tools? Agentic tools can plan and execute multi-step tasks with limited human intervention, such as implementing a feature across several files, running tests, and fixing failures, rather than just suggesting one line or function at a time. The key distinction is autonomy over a sequence of actions, not just the sophistication of a single suggestion.
Should we standardize on one AI coding tool company-wide, or let teams choose? Most engineering leaders who’ve gone through this find a hybrid approach works best: a small, vetted set of approved tools (often two or three) rather than one mandated tool or fully unrestricted choice. This balances developer preference and task fit against the real overhead of supporting, securing, and licensing too many overlapping tools.
How should code review change for AI-generated code? Treat AI-assisted pull requests with the same or greater scrutiny as human-written ones, with specific attention to subtle logic errors, security issues, and license or provenance concerns that AI-generated code can introduce in ways that look superficially correct. Some organizations add a distinct review checklist item flagging which parts of a change were AI-generated.
Are no-code, natural-language app builders a threat to professional engineering teams? They’re better understood as a new entry point for software creation, particularly for prototypes, internal tools, and early-stage products, rather than a replacement for professional engineering. The practical question for engineering leaders is establishing a clear path for when and how software built this way graduates into a properly engineered, supported codebase.
Do we need a separate framework for building our own AI agents, or can our coding assistant handle that too? These are typically different tools solving different problems. Coding assistants help developers write and modify code faster. Agent orchestration frameworks help developers build AI agents that are part of a product itself, with their own reasoning, tool use, and decision-making. Organizations building AI features into their own products generally need both.
This article is part of the SD Times 100 2026 series exploring the categories and companies shaping software development this year. Read the full SD Times 100 2026 list for the complete roundup.
