AI Coding Market Map
Agents

IDE-Native Coding Agents

Nascent
Emerging
Growth
Maturing
Mature

AI coding tools that live inside your editor — from autocomplete extensions to fully autonomous agent-native IDEs. These are the tools most engineering teams encounter first, and understanding the split between copilot-style extensions and agentic IDEs is critical to making the right investment.

Overview Recommendations Trends and Strategic Signals Tools Adoption and Traction New Entrants & Watch List

Overview

Category maturity: Growth. The IDE-native AI tools category has moved decisively past experimentation into broad organizational adoption — Cursor at $2B ARR, Copilot code reviews reaching 60 million, half of Fortune 500 reportedly using Cursor. The transition from autocomplete assistance to always-on agentic automation marks the next structural evolution, though that capability layer is still in early rollout.

Direction of travel: Over the next 6-12 months, the category will shift along three axes:

  • The dominant interface will move from manual prompting to event-driven orchestration, with IDE agents acting on incoming signals (code changes, tickets, incidents) rather than waiting for developer input
  • Distribution will flow through protocol-level standards like ACP and MCP, making agent interoperability across IDEs table stakes
  • Enterprise differentiation will concentrate on governance, data residency, and team-level orchestration rather than raw code generation quality, which is converging across tools

Coalesced patterns: The foundations are well established:

  • Multi-model selection (OpenAI, Anthropic, Google), inline completions, multi-file agentic edits, and in-editor chat are standard across most tools
  • MCP integration for connecting agents to external systems (databases, CI/CD, ticketing) is quickly becoming a baseline expectation
  • Pricing has settled around per-seat monthly subscriptions with tiered credit systems for heavier model usage
  • VS Code extensions remain the dominant distribution mechanism, with JetBrains plugins increasingly on par for enterprise polyglot teams

Unsolved problems: Several gaps remain open:

  • Enterprise governance of agentic output — audit trails for agent-generated commits, rollback procedures, attribution — is poorly standardized and largely left to teams to solve
  • The shift to always-on automation introduces new risk categories, including agents making unintended changes without sufficient human review; no tool has fully addressed this at the policy layer
  • Cost predictability at team scale is a friction point, as credit-based pricing makes monthly spend difficult to forecast
  • Context management at large codebase scale remains genuinely hard; most tools perform well on focused tasks but lose coherence across very large or highly modular repositories

Recommendations

1. Cursor is the right default starting point for teams targeting a software factory model. Automations (launched March 5) is the first IDE-native tool to deliver event-driven, always-on agent operation at team scale: agents that trigger on code changes, Slack messages, and incidents rather than waiting for developer prompts. At $2B ARR and $40/user/month at the Business tier, it has the adoption base, financial stability, and governance features for confident enterprise deployment. If you are not yet on Cursor Business, now is the right moment to move.

2. The organizational work is the constraint, not the tooling. The shift to factory-model operation requires three things the tools do not provide: clear specs that agents can act on without constant clarification, a PR review process that can handle agent-generated output volume without becoming a bottleneck, and a policy for attribution and rollback when agents introduce regressions. Address these before expanding Automations use; the teams that will capture the most leverage are those that redesign the development workflow alongside the tool deployment.

3. The GitHub Copilot data policy change requires an immediate team audit. Starting April 24, individual Copilot plans (Free, Pro, Pro+) default to sharing interaction data with GitHub for model training. Any developer using these tiers is affected unless they opt out at /settings/copilot/features. Ensure developers working with sensitive code are on Business or Enterprise and document the policy position before April 24.

4. Watch Augment Code's Intent for the next evolution of factory architecture. Intent's spec-driven multi-agent model (coordinator plans, implementors execute in parallel, verifier checks outputs against spec) is the closest current expression of a true factory assembly model. It is macOS-only today, limiting team-wide deployment, but Windows availability would make it worth a structured evaluation. Track it as an indicator of where the category is heading rather than a deployment candidate for most teams right now.

1. Agentic automation crosses the "always-on" threshold. The most consequential shift this week is Cursor's March 5 launch of Automations, a system that triggers AI coding agents from external events: a Slack message, a PagerDuty incident, a timer, or a new code commit. This moves the dominant workflow from "human prompts, AI responds" to "AI runs continuously, human reviews outputs." Cursor reports running hundreds of automations per hour internally, and customers are already using it for security audits, incident response, and codebase summaries. The category is accelerating past on-demand assistance into ambient engineering infrastructure.

2. JetBrains' Agent Client Protocol is reshaping distribution. JetBrains' ACP standard, with Cursor joining the registry in early March, establishes a meaningful new distribution channel for AI coding agents. JetBrains IDEs serve a large portion of enterprise Java and polyglot developers who previously had limited AI-native options. The ACP Registry now lets developers install AI agents in one click, and tools like Claude Agent, Codex, and Cursor are all live. For enterprise teams standardized on IntelliJ or PyCharm, the barrier to adopting frontier AI agents has dropped substantially. Teams on JetBrains IDEs no longer need to choose between their editor and factory-model agents — this removes the last major friction point for factory-model adoption in Java and Kotlin organizations.

3. GitHub Copilot's data policy change is the week's sharpest enterprise signal. Starting April 24, GitHub will use interaction data from Copilot Free, Pro, and Pro+ users to train AI models by default, with an opt-out available at /settings/copilot/features. Business and Enterprise tiers are explicitly excluded. This will sharpen the purchasing conversation for enterprise buyers: teams with code confidentiality requirements now have a clear reason to upgrade to Business or Enterprise rather than rely on individual-tier plans, and compliance officers at regulated companies will want to confirm their developers are on the right tier.

4. Google Gemini Code Assist went free, adding a well-resourced competitor to the IDE extension tier. In March 2026, Google made Gemini Code Assist fully free for individual developers, offering VS Code and JetBrains plugins powered by Gemini 2.5 with 6,000 code requests and 240 chat requests daily. The free tier includes inline completions, multi-file generation, and unit test generation. This raises the floor for what "free" means in the category and puts direct pressure on freemium tiers from Cline, Continue.dev, and Copilot Free. The catch: free-tier interactions are used to improve Google's models, a data consideration for enterprise code.

5. Windsurf's parallel agent architecture and competitive benchmarks put it credibly ahead of Cursor in developer power rankings. Windsurf's Wave 13 release delivered Arena Mode (side-by-side model comparison with blind voting), Plan Mode (structured task planning before code execution), and first-class parallel multi-agent sessions via Git worktrees. LogRocket's March 2026 AI Dev Tool Power Rankings placed Windsurf #1, ahead of Cursor and Copilot. Under Cognition AI ownership since December 2025, the product is shipping faster than during its OpenAI acquisition talks, and its 1M+ active users and $82M ARR at acquisition provide a stable foundation.

Tools

Agent-Native IDEs — standalone IDEs or platforms with deep agentic capabilities: event-driven automation, multi-agent coordination, spec-driven workflows, or autonomous background agents.

Cursor logoCursor

  • Maker: Anysphere
  • Model(s): GPT-4.1, Claude Sonnet/Opus, Gemini 2.5 Pro, and others selectable; Cursor-native model available
  • Access: Free tier; Pro at $20/month; Business at $40/user/month; Enterprise pricing available. cursor.com
  • Strengths:
    • Automations platform (launched March 2026) enables event-triggered agents that run continuously on code changes, Slack signals, timers, and external incidents, moving the workflow from on-demand to ambient.
    • $2B ARR and $29.3B valuation demonstrate category leadership and financial stability for enterprise procurement conversations.
    • Now available inside JetBrains IDEs via ACP, extending reach to enterprise Java and polyglot teams without requiring an IDE switch.
    • Self-hosted cloud agents (announced March 2026) keep code and tool execution within customer infrastructure, addressing a key enterprise data residency requirement.
  • Limitations:
    • Teams deploying Automations should plan for governance policies around agent-generated commits and review workflows; this capability outpaces current out-of-box guardrails.
    • Business tier provides the right access controls for team deployment; individual Pro plans carry standard cloud data handling terms that compliance teams should review.
    • The breadth of model choice and configuration options benefits power users; teams that want simpler, more opinionated setups may find the surface area larger than needed at initial rollout.
  • Enterprise readiness: Enterprise-ready. SSO, audit logs, admin controls, and self-hosted agent execution available at Business and Enterprise tiers; SOC 2 compliance in place.
  • Software factory readiness: Leading. Automations delivers event-triggered, always-on agent operation today; self-hosted agent execution, MCP integrations, and $2B ARR adoption base make Cursor the most mature factory-model tool in this landscape. The governance layer (agent commit attribution, rollback policy) is the remaining gap teams need to solve themselves.
  • Best for: Engineering organizations ready to move from individual AI assistance to team-wide agentic automation, particularly those with complex multi-file codebases and event-driven development workflows.
  • This week: Automations launch (March 5) received broad coverage as the most significant agentic capability expansion in the category. JetBrains ACP availability expanded the addressable enterprise audience. Self-hosted cloud agent support announced.

Windsurf logoWindsurf (Codeium / Cognition AI)

  • Maker: Codeium, acquired by Cognition AI (December 2025)
  • Model(s): Cascade (proprietary), Claude, GPT-4.1, Gemini, and others; Arena Mode enables side-by-side model comparison
  • Access: Free tier; Pro at $15/month; Teams available; Max plan recently launched. windsurf.com
  • Strengths:
    • Wave 13 Arena Mode enables blind side-by-side model comparison in real time, giving teams empirical data on which model performs best for their specific codebase and tasks.
    • Plan Mode structures complex tasks into reviewable plans before any code is written, reducing agent drift on multi-step work.
    • First-class parallel multi-agent sessions with Git worktrees and side-by-side Cascade panes enable true concurrent development approaches.
    • Ranked #1 in LogRocket AI Dev Tool Power Rankings (March 2026), demonstrating strong developer sentiment relative to better-funded competitors.
  • Limitations:
    • Enterprise governance documentation is still maturing post-acquisition; teams should validate current SOC 2 status and data handling terms under Cognition AI ownership.
    • The acquisition transition (December 2025) means procurement and support escalation paths are still stabilizing; enterprise buyers should confirm SLA commitments before large-scale rollout.
    • The free tier's generous access relies on credit quotas; teams with heavy agentic use should model credit consumption carefully before committing to team plans.
  • Enterprise readiness: Developing. Enterprise tier available with SSO and admin controls; post-acquisition compliance documentation should be verified before regulated-industry deployments.
  • Software factory readiness: Advancing. Parallel multi-agent sessions, Plan Mode, and Git worktree support enable concurrent factory-style workstreams today. Event-triggered automation (the defining factory-model capability) is not yet at Cursor's level; Windsurf currently excels at parallelizing human-initiated work rather than eliminating the human trigger entirely.
  • Best for: Developer teams that want the most advanced agentic IDE capabilities available today and are willing to be on the leading edge of multi-agent workflow patterns.
  • This week: Wave 13 features (Arena Mode, Plan Mode, parallel agents) confirmed as shipped. Max plan pricing added. Windsurf holds #1 position in March 2026 developer power rankings.

Kiro logoKiro

  • Maker: Amazon Web Services
  • Model(s): Claude Sonnet 4.5 and 4.0, with auto model selection
  • Access: Generally available (November 17, 2025); free tier available; enterprise pricing through AWS. kiro.dev
  • Strengths:
    • The only IDE in this landscape with built-in spec-driven development: requirements.md with EARS notation, design.md with technical architecture, and implementation tasks are generated within the IDE before any code is written
    • Agent Hooks (event-driven automation) trigger automated spec compliance checks, security scans, and test suite runs on filesystem changes, directly addressing auditability and continuous governance
    • AWS ecosystem depth: SageMaker Studio remote connectivity (March 2026) and HealthOmics extension (January 2026) show active platform expansion for enterprise data and ML teams
    • Built on Code OSS with support for VS Code extensions, themes, and settings, reducing migration friction for teams already on VS Code
  • Limitations:
    • As a proprietary IDE, teams adopting Kiro accept a dependency on AWS and the Kiro product roadmap; teams already invested in Cursor or VS Code will face a migration cost
    • Enterprise pricing is not publicly listed, which complicates budget planning; plan for a sales-cycle conversation with AWS
    • Kiro works with any technology stack but its deepest integrations and examples are AWS-native; teams on GCP or Azure will get less native workflow support
  • Enterprise readiness: Enterprise-ready. General availability, institutional AWS backing, formal spec methodology, agent hooks for governance, and active enterprise integrations.
  • Software factory readiness: Advancing. Spec-driven development with automated compliance hooks is the most structured approach to factory-model governance in any IDE today. The spec layer gives agents durable context that survives across sessions, addressing one of the key failure modes in factory-scale operation.
  • Best for: Engineering teams in the AWS ecosystem who want a spec-first IDE with built-in governance hooks and native cloud service connectivity.
  • This week: SageMaker Studio remote connectivity from Kiro and Cursor announced in March 2026, deepening the platform's ML/data engineering value proposition.

Augment Code logoAugment Code

  • Maker: Augment Code (founded 2022; $252M raised)
  • Model(s): Multiple frontier models via Augment Context Engine; model-agnostic at the agent layer
  • Access: IDE extension with paid team plans; Intent macOS desktop app in public beta (uses Augment credits). augmentcode.com
  • Strengths:
    • Context Engine provides persistent, codebase-wide understanding that informs every completion and agent action, making suggestions more accurate on large or complex repositories than generic RAG-based approaches.
    • MCP integration (launched February 2026) connects agents to deployment pipelines, debug tools, and external infrastructure, enabling end-to-end task execution beyond the IDE.
    • Intent macOS desktop app (February 2026 public beta) provides a dedicated multi-agent orchestration workspace where a coordinator agent fans work to implementor agents based on a living spec, then verifies results; this is a differentiated approach to spec-driven development.
    • $252M in funding from tier-1 investors (Sutter Hill, Index, Lightspeed) provides a stable financial foundation relative to earlier-stage competitors.
  • Limitations:
    • Intent is currently macOS-only (Windows on waitlist), limiting team-wide deployment for organizations with mixed operating system environments.
    • The Augment IDE extension and Intent are complementary products with distinct interfaces; teams should clarify which workflow best fits their use case before rollout to avoid adoption confusion.
    • Developer community visibility and public traction signals (GitHub stars, forum activity) are lower than Cursor or Cline; teams evaluating on community ecosystem strength should factor this in.
  • Enterprise readiness: Developing. Enterprise-grade context engine and compliance posture maturing; audit and governance features appropriate for growth-stage teams; verify current SOC 2 status for regulated-industry procurement.
  • Software factory readiness: Advancing. Intent's spec-driven multi-agent orchestration (coordinator, implementors, verifier) is architecturally the closest thing in this landscape to a factory assembly model. The Context Engine provides the codebase understanding needed for agents to operate reliably on large scopes without constant human correction. macOS-only limitation constrains full team deployment today; watch for Windows availability as a trigger for broader rollout evaluation.
  • Best for: Engineering teams working on large, complex codebases where codebase-wide context quality is the bottleneck, and organizations interested in spec-driven multi-agent development workflows.
  • This week: Intent macOS multi-agent orchestration workspace confirmed in public beta. MCP support for agent-to-infrastructure connections added in February 2026. These are the most significant capability expansions for Augment in this period.
bg-gray-800/50 rounded-lg select-none"> Copilot Extensions Included for completeness — agentic solutions are the stronger investment for teams scaling AI-assisted development.

GitHub Copilot logoGitHub Copilot

  • Maker: Microsoft / GitHub
  • Model(s): GPT-4.1 (now included on paid plans), GPT-5 mini, Claude (selectable), Gemini (selectable)
  • Access: Free tier (limited); Pro at $10/month (individual); Business at $19/user/month; Enterprise at $39/user/month. github.com/features/copilot
  • Strengths:
    • Native GitHub integration provides the most complete context for code review, pull request automation, and issue-to-PR agent workflows of any tool in this landscape.
    • Copilot code review reached 60 million reviews by March 2026, demonstrating at-scale, proven deployment across development teams.
    • GPT-4.1 is now included on paid plans without consuming premium request quota, improving cost predictability at team scale.
    • Business and Enterprise tiers provide the strongest current compliance posture in this landscape: no interaction data used for model training, dedicated admin controls, and clear data handling boundaries.
  • Limitations:
    • The April 24 policy change makes the distinction between individual and team tiers newly important: Free, Pro, and Pro+ users' interaction data will be used for model training by default unless opted out, which is a meaningful compliance consideration for teams that have not moved to Business or Enterprise.
    • Agentic capabilities (agent mode, multi-file edits) are maturing and functional but trail Cursor and Windsurf in depth of orchestration and automation trigger options at current state.
    • Multi-model selection, while available, requires configuration; teams accustomed to a more opinionated single-model experience may need adjustment time.
  • Enterprise readiness: Enterprise-ready. SSO, audit logs, IP indemnification, data privacy guarantees (no training on Business/Enterprise data), and SLA support all available. The strongest enterprise compliance posture in the category.
  • Software factory readiness: Developing. Issue-to-PR agent assignment and Copilot code review at 60M reviews represent meaningful factory-adjacent automation within the GitHub workflow. Factory-scale event-driven operation inside the IDE trails Cursor; Copilot's factory model strength is in the GitHub Actions and PR pipeline layer rather than the editor layer.
  • Best for: Engineering organizations already in the GitHub ecosystem that want team-scale deployment with the lowest compliance risk and the tightest issue-to-PR workflow integration.
  • This week: Privacy policy update announced March 25, effective April 24: Free, Pro, and Pro+ interaction data used for model training by default with opt-out. GPT-4.1 added to paid plans without premium quota consumption. This is the most important enterprise signal from Copilot this week.

Amazon Q Developer logoAmazon Q Developer

  • Maker: Amazon Web Services
  • Model(s): Amazon Q (proprietary); integrated with Amazon Bedrock model ecosystem
  • Access: Free tier (limited); Pro at $19/user/month. IDE plugins available for VS Code, JetBrains, Visual Studio, Eclipse. aws.amazon.com/q/developer
  • Strengths:
    • Deepest AWS-native integration of any tool in this landscape: direct access to CloudFormation, Lambda, CodePipeline, and IAM policies from within the IDE makes it the natural choice for teams building on AWS infrastructure.
    • Agent capabilities for documentation generation, code reviews, and unit test generation are mature and benchmarked, with SWE-Bench Verified performance at 49% as of late 2025.
    • Eclipse IDE plugin reached general availability in June 2025, completing broad coverage across VS Code, JetBrains, Visual Studio, and Eclipse.
    • Enterprise data handling is governed by AWS's existing compliance framework, which is already in place for most enterprise AWS customers.
  • Limitations:
    • Teams not primarily building on AWS will find the tool's differentiated value narrower than its AWS-centric positioning; the integration advantages are most pronounced within AWS workflows.
    • Agentic workflow automation at the level of Cursor's Automations is not yet available; the tool excels at task-level assistance within the editor rather than event-driven ambient operation.
    • Community momentum and developer mindshare trail Cursor and Copilot; teams evaluating on the basis of ecosystem energy and third-party integration depth should weight this accordingly.
  • Enterprise readiness: Enterprise-ready. Governed by AWS compliance infrastructure including SOC 2, HIPAA, FedRAMP. Data privacy and admin controls well-documented. Strong SLA support through AWS enterprise agreements.
  • Software factory readiness: Limited. Amazon Q's factory-model value concentrates in AWS-native workflow automation (CloudFormation, CodePipeline, Lambda) rather than IDE-level event-driven agent operation. Teams building a software factory on non-AWS infrastructure will find limited leverage here.
  • Best for: Engineering teams with significant AWS infrastructure investment who want AI coding assistance with native cloud service context and AWS-level compliance guarantees.
  • This week: No new signals this week; carrying forward prior profile. Eclipse GA (June 2025) remains the most recent notable expansion.

JetBrains AI Assistant logoJetBrains AI Assistant

  • Maker: JetBrains
  • Model(s): Multiple selectable via model selector; supports Bring Your Own Key for third-party providers; optimized for JetBrains-native IDE context
  • Access: Included with JetBrains IDE subscriptions; AI Pro available as add-on. jetbrains.com/ai-assistant
  • Strengths:
    • The 2026.1 EAP introduces Recap and Insights, two experimental features that proactively surface recent activity summaries and non-obvious code explanations, reducing context-switching overhead for developers resuming complex work.
    • ACP Registry makes JetBrains IDEs the most open AI agent ecosystem available: Cursor, Claude Agent, Codex, and GitHub Copilot are all accessible in one-click installation alongside the native AI Assistant.
    • Quota-free next edit suggestions (now shipping in 2026.1) propagate changes across a file without consuming premium credits, benefiting teams with high edit frequency.
    • Multi-file edits and web search from chat (/web) are now available directly in the AI chat panel.
  • Limitations:
    • The AI Assistant's native agentic capabilities are still developing relative to Cursor or Windsurf; its primary strength is deep IDE-native context and the ACP ecosystem rather than autonomous multi-step orchestration.
    • Teams that want to stay within a single, vertically integrated AI coding product will find the ACP model requires managing agent credentials and configurations across multiple tools.
    • Experimental features (Recap, Insights) require installing a separate plugin and are in early access; production-grade availability should be confirmed before broad team rollout.
  • Enterprise readiness: Enterprise-ready. JetBrains' existing enterprise licensing, SSO, and data handling policies apply; Bring Your Own Key support allows teams to route through their own model agreements for stricter data control.
  • Software factory readiness: Developing via ACP. The JetBrains AI Assistant itself is an IDE-assistance product, not a factory orchestration tool. Its factory-model relevance comes from the ACP Registry: teams on JetBrains IDEs can now connect Cursor Automations or Claude Agent and inherit their factory capabilities without switching editors. Evaluate JetBrains AI as a platform access layer rather than a factory engine.
  • Best for: Teams already standardized on JetBrains IDEs for Java, Kotlin, Python, or polyglot enterprise development, particularly those who want to preserve IDE investment while accessing frontier AI agents via ACP.
  • This week: 2026.1 EAP released with Recap and Insights experimental features. ACP Registry now lists Cursor as an available agent. Quota-free next edit suggestions confirmed shipping. These are the most substantive JetBrains AI updates in several months.

Cline logoCline

  • Maker: Community / open source (cline.bot)
  • Model(s): Bring your own: OpenAI, Anthropic, Google, Ollama, and any API-compatible endpoint
  • Access: Free and open source (VS Code Marketplace); Teams plan at $20/month (first 10 seats always free, free through Q1 2026). cline.bot
  • Strengths:
    • 5M+ installs makes it the most widely adopted open-source AI coding agent in the VS Code ecosystem, with a large community producing integrations, guides, and workflow patterns.
    • Plan/Act mode separation gives developers explicit control over when the agent plans and when it executes, addressing a key risk concern with autonomous coding agents.
    • MCP Marketplace integration enables connections to external tools and infrastructure directly from the agent's workflow.
    • Full transparency into agent actions (every file operation and command requires user confirmation by default) makes it well-suited to teams that want auditability without a commercial enterprise agreement.
  • Limitations:
    • Bring-your-own-model architecture means teams are responsible for managing API keys, rate limits, and cost tracking independently; enterprise-grade cost visibility requires additional tooling.
    • Teams plan pricing transitions from free after Q1 2026 at $20/month; teams that built workflows on the free tier should confirm pricing before broad rollout.
    • Governance features (audit logs, centralized policy controls, SSO) depend on the team's own API provider and infrastructure rather than being built into Cline itself; enterprise compliance requires self-managed policy implementation.
  • Enterprise readiness: Early. Well-suited to team adoption via self-managed policies; enterprise governance runs through API provider agreements and internal controls at current maturity.
  • Software factory readiness: Developing with effort. Plan/Act separation and MCP integrations are the right building blocks for factory workflows, but Cline lacks a native event-trigger layer. Teams willing to wire their own orchestration (e.g., triggering Cline from CI events or webhooks) can approach factory operation; teams that want this out of the box should look to Cursor Automations first.
  • Best for: Engineering teams that want maximum model flexibility and full operational transparency at low cost, particularly those comfortable managing their own LLM API relationships and willing to build internal governance practices.
  • This week: No new product signals this week. Teams plan pricing confirmation (free through Q1 2026, $20/month thereafter) is the most relevant near-term planning item.

Continue.dev logoContinue.dev

  • Maker: Continue (open source; continuedev/continue on GitHub)
  • Model(s): Bring your own: OpenAI, Anthropic, Google, Ollama, LM Studio, and any API-compatible or locally hosted model
  • Access: Free and open source. continue.dev
  • Strengths:
    • The most model-agnostic tool in this landscape: works with any API-compatible endpoint including local models via Ollama or LM Studio, making it uniquely suited to air-gapped or on-premises deployments.
    • MCP support connects AI workflows to external systems (databases, documentation, CI/CD) without requiring a commercial vendor agreement.
    • Available in both VS Code and JetBrains, supporting teams that have heterogeneous editor environments.
    • Automated AI checks enforceable in CI (now a core feature per the GitHub repo positioning) enable workflow-level quality gates beyond the IDE.
  • Limitations:
    • No commercial support tier currently; teams that need SLA-backed support or vendor accountability will need to source those through their model provider rather than Continue itself.
    • UX and agentic orchestration polish trail commercial alternatives; Continue rewards developer investment in configuration and customization rather than delivering an out-of-box experience.
    • Community-driven development means release cadence and roadmap visibility are less predictable than commercial products; teams building production workflows should monitor the GitHub release cadence closely.
  • Enterprise readiness: Early. Ideal for self-managed, on-premises, or air-gapped environments where model control and data residency are paramount; enterprise governance is fully self-managed.
  • Software factory readiness: Limited. Continue's CI-enforceable AI checks are a useful factory building block, and local model support offers cost control at scale, but the tool is optimized for individual developer control rather than team-level orchestration. Better positioned as a component in a self-built factory stack than as a primary factory tool.
  • Best for: Teams that require full control over model selection and data routing, including those running local models for compliance reasons or operating in restricted network environments.
  • This week: No new signals this week.

Tabnine logoTabnine

  • Maker: Tabnine
  • Access: Code Assistant at $39/user/month; Agentic Platform at $59/user/month. tabnine.com
  • Note: Tabnine's primary differentiation is privacy-by-design architecture and on-premises deployment, making it the right tool for regulated industries (financial services, healthcare, defense) with strict data sovereignty requirements. These conditions are not present in most growth-stage B2B SaaS portfolios, so Tabnine is not a primary recommendation here. Teams at companies selling into regulated verticals who need to demonstrate strong code security posture to enterprise customers may find the positioning useful. Not profiled in depth in this landscape; evaluate when regulated-industry deployment requirements are confirmed.

Adoption and Traction

  • Cursor: $2B ARR as of early March 2026, doubling in roughly three months. $29.3B valuation from Series D. Half of Fortune 500 reportedly using Cursor. Company-wide rollout at Money Forward, with engineers saving 15-20 hours per week. Automations launch covered by TechCrunch, Dataconomy, and multiple tech outlets.

  • Windsurf: Ranked #1 in LogRocket AI Dev Tool Power Rankings (March 2026), ahead of Cursor and Copilot. 1M+ active users; $82M ARR at December 2025 Cognition acquisition. Wave 13 features (Arena Mode, Plan Mode, parallel agents) shipped in February 2026.

  • GitHub Copilot: Copilot code reviews reached 60 million as of March 2026, growing 10x since April 2025 launch. GPT-4.1 now included on paid plans without consuming premium request quota. Data policy change generating significant press and developer community discussion.

  • JetBrains AI Assistant: 2026.1 EAP shipped with Recap and Insights experimental features. ACP Registry live with Cursor, Claude Agent, and Codex available in one-click install. ACP standard gaining developer mindshare as a neutral interoperability layer.

  • Augment Code: Intent macOS desktop agent launched in public beta (February 2026); MCP support added for agent-to-infrastructure connections. Total funding: $252M. Windows access on a waitlist.

  • Kiro: General availability since November 17, 2025; SageMaker Studio remote connectivity added March 2026; AWS HealthOmics extension launched January 2026. The only IDE with built-in spec-driven development workflow.

  • Google Gemini Code Assist: Free tier launched March 2026 with VS Code and JetBrains plugins powered by Gemini 2.5. Covered across developer publications as a meaningful market entry.

New Entrants & Watch List

Google Gemini Code Assist (IDE-native; recommend adding to core coverage)

  • Maker: Google
  • Description: VS Code and JetBrains plugins powered by Gemini 2.5, made fully free for individual developers in March 2026. Offers inline completions, multi-file generation, unit test generation, and a context-aware chat interface. 6,000 code requests and 240 chat requests daily on the free tier. Enterprise Standard and Enterprise tiers available with Google Cloud compliance frameworks.
  • Why it warrants watching: Google's free individual tier raises the floor for what "free" means in the category and could accelerate bottom-up adoption in organizations already using Google Cloud. Gemini 2.5 model quality is competitive with GPT-4.1 on coding benchmarks. Enterprise tiers carry Google Cloud's compliance posture, which is already in procurement scope for many B2B SaaS companies. The free-tier data-for-training tradeoff (interactions used to improve Google models) mirrors the Copilot Free pattern and creates an identical enterprise tier upgrade conversation. Recommend adding to core coverage in the next weekly cycle given Google's distribution reach and enterprise relevance.

Note on Augment Intent: Augment Code's Intent macOS desktop application functions as a multi-agent orchestration workspace rather than a primary IDE extension. It connects to IDE tools and agents (including Claude Code and Codex) but is itself a standalone macOS app. This product sits at the boundary of the IDE-native and cloud-hosted agent categories; future weekly entries will track it in the Watch List until its IDE integration posture clarifies.

← Previous CLI Agents Agents Next → Cloud Agents Agents