Where It Fits
Where does Devin Desktop (formerly Windsurf) belong in the stack?
Devin Desktop (formerly Windsurf)'s primary role: Agent Editor & Agent Command Center. It is Core across Build, supports Plan and Validate, and touches Ship and Operate.
Plan
- Work planning: Supports
Build
- Codebase context: Core
- Coding: Core
- Agent runtime: Core
- Tool execution: Supports
Validate
- Debugging: Supports
- Testing: Supports
- Code review: Supports
- Application security: Touches
Ship
- Source control: Touches
- CI/CD: Touches
Operate
- Observability: Touches
Devin Desktop (formerly Windsurf) beside the rest of its stack
Every product we cover on Agentic Software Development, on the layers it owns, placed by what its practitioners report. Pick any of them to compare.
Signals from
Show products that
Activating a product opens its read here. Activating the selected product opens its hub.
Each product sits on the layers it owns, placed left to right by how far its positive sources lead its negative ones. A disc's area is its sources in this view and its tone is its band. Products with fewer than 20 sources from 15 authors have no net and sit at the right, under Unscored. The teal line joins Devin Desktop (formerly Windsurf) across its layers.
Read the field as a table
| Product | Core on | Supports | Overall Signal | Net | Sources | Positive | Mixed | Negative |
|---|---|---|---|---|---|---|---|---|
| Claude Code | Codebase context, Coding, Agent runtime, Tool execution | Work planning, Debugging, Testing, Code review | Strong positive | +75 | 53 | 41 | 11 | 1 |
| v0 | Coding, Agent runtime | Work planning, Codebase context, Tool execution, Debugging, Source control, CI/CD | Strong positive | +68 | 24 | 19 | 3 | 2 |
| Bolt | Coding, Agent runtime | Work planning, Codebase context, Tool execution, Debugging, Application security, Source control, CI/CD | Strong positive | +60 | 20 | 14 | 4 | 2 |
| Codex | Codebase context, Coding, Agent runtime, Tool execution | Work planning, Debugging, Testing, Code review | Strong positive | +55 | 33 | 21 | 10 | 2 |
| Linear | Work planning | Codebase context, Coding, Agent runtime, Tool execution, Code review, Source control | Positive | +41 | 103 | 57 | 33 | 13 |
| Greptile | Code review | Codebase context, Tool execution, Testing, Application security, Source control | Positive | +40 | 20 | 11 | 6 | 3 |
| Devin | Coding, Tool execution | Work planning, Codebase context, Agent runtime, Debugging, Testing, Code review, Application security | Positive | +39 | 24 | 13 | 7 | 4 |
| Lovable | Coding, Agent runtime | Work planning, Codebase context, Tool execution, Debugging, Application security, Source control, CI/CD | Positive | +27 | 30 | 12 | 14 | 4 |
| Qodo | Code review | Codebase context, Tool execution, Application security, Source control | Mixed | +17 | 23 | 9 | 9 | 5 |
| CodeRabbit | Code review | Work planning, Codebase context, Coding, Tool execution, Testing, Application security, Source control | Mixed | +16 | 46 | 23 | 9 | 14 |
| GitHub Copilot | Coding, Agent runtime, Code review | Work planning, Codebase context, Tool execution, Debugging, Testing, Application security, Source control | Mixed | +15 | 20 | 7 | 9 | 4 |
| Cursor | Codebase context, Coding, Agent runtime | Work planning, Tool execution, Debugging, Testing, Code review | Mixed | +11 | 47 | 19 | 14 | 14 |
| Seer | Debugging, Observability | Codebase context, Coding, Tool execution, Testing, Code review | Mixed | +11 | 29 | 10 | 13 | 6 |
| Replit Agent | Coding, Agent runtime, CI/CD | Work planning, Codebase context, Tool execution, Debugging, Testing, Application security, Source control, Observability | Below minimum | none | 18 | 15 | 3 | 0 |
| Junie | Codebase context, Coding, Agent runtime | Work planning, Tool execution, Debugging, Testing, Code review | Below minimum | none | 15 | 9 | 4 | 2 |
| Devin Desktop (formerly Windsurf) | Codebase context, Coding, Agent runtime | Work planning, Tool execution, Debugging, Testing, Code review | Below minimum | none | 14 | 7 | 5 | 2 |
| Google Antigravity | Coding, Agent runtime, Tool execution | Work planning, Codebase context, Debugging, Testing, Code review | Below minimum | none | 13 | 3 | 0 | 10 |
| OpenCode | Coding, Agent runtime, Tool execution | Work planning, Codebase context, Debugging, Testing | Below minimum | none | 13 | 12 | 1 | 0 |
| Gemini CLI | Coding, Agent runtime, Tool execution | Work planning, Codebase context, Debugging, Testing, Code review, CI/CD | Below minimum | none | 12 | 1 | 8 | 3 |
| Zed | Coding, Agent runtime | Codebase context, Tool execution, Debugging | Below minimum | none | 11 | 7 | 2 | 2 |
| Auggie CLI | Codebase context, Coding, Agent runtime | Work planning, Tool execution, Debugging, Testing, Code review | Below minimum | none | 8 | 5 | 2 | 1 |
| Cline | Coding, Agent runtime, Tool execution | Work planning, Codebase context, Debugging, Testing | Below minimum | none | 8 | 5 | 2 | 1 |
| Amazon Q Developer | Coding, Agent runtime | Codebase context, Tool execution, Debugging, Testing, Code review, Application security, Observability | Below minimum | none | 6 | 1 | 2 | 3 |
| Amp | Coding, Agent runtime | Work planning, Codebase context, Tool execution, Debugging, Testing, Code review, Source control | Below minimum | none | 6 | 5 | 0 | 1 |
| Continue | Coding, Agent runtime | Codebase context, Tool execution, Debugging, Testing | Below minimum | none | 6 | 1 | 5 | 0 |
| Qwen Code | Coding, Agent runtime, Tool execution | Work planning, Codebase context, Debugging, Testing, Code review | Below minimum | none | 6 | 2 | 1 | 3 |
| T3 Code | Agent runtime | Work planning, Coding, Tool execution, Source control | Below minimum | none | 5 | 4 | 0 | 1 |
| Factory | Coding, Agent runtime, Tool execution | Work planning, Codebase context, Debugging, Testing, Code review, Application security | Below minimum | none | 4 | 1 | 2 | 1 |
| Aider | Codebase context, Coding, Agent runtime | Work planning, Tool execution, Debugging, Testing, Source control | Below minimum | none | 3 | 0 | 3 | 0 |
| OpenHands | Coding, Agent runtime, Tool execution | Work planning, Codebase context, Debugging, Testing, Code review, Application security | Below minimum | none | 3 | 2 | 1 | 0 |
| Jules | Coding, Tool execution | Work planning, Codebase context, Agent runtime, Debugging, Testing | Below minimum | none | 2 | 1 | 1 | 0 |
| Kiro | Coding, Agent runtime | Work planning, Codebase context, Tool execution, Debugging, Testing, Code review | Below minimum | none | 2 | 2 | 0 | 0 |
| Warp | Coding, Agent runtime, Tool execution | Work planning, Codebase context, Debugging, Testing, Code review, Source control | Below minimum | none | 2 | 1 | 1 | 0 |
| Kimi Code | Coding, Agent runtime, Tool execution | Work planning, Codebase context, Debugging, Testing | Below minimum | none | 1 | 1 | 0 | 0 |
| Mistral Vibe | Coding, Agent runtime, Tool execution | Work planning, Codebase context, Debugging, Testing | Below minimum | none | 1 | 0 | 1 | 0 |
| Tabnine | Codebase context, Coding | Agent runtime, Tool execution, Debugging, Testing, Code review | Below minimum | none | 1 | 0 | 0 | 1 |
| kcap | Codebase context | Agent runtime, Tool execution, Code review | Below minimum | none | 0 | 0 | 0 | 0 |
Each author counts once.
Stack Footprint
What is Devin Desktop (formerly Windsurf) used for, and by whom?
AI Coding Agents, also listed under AI development environment, developer tools, and software development.
Plan
Work planning: Supports
Work planning, supports. Devin Local's Plan mode researches the codebase and writes an editable plan before implementing, and Spaces group sessions by task; planning stays in the team's tracker.
Build
Codebase context: Core
Codebase context, core. Devin Desktop indexes the entire local codebase for retrieval, adds a Fast Context subagent for code search, and on Enterprise plans its plugins can index remote repositories.
Coding: Core
Coding, core. Devin Desktop's primary role is coding implementation: changing repository code with Devin Local or an ACP agent inside the editor.
Agent runtime: Core
Agent runtime, core. The editor and its Agent Command Center are where developers start, track, and review local agents, third-party ACP agents, and Devin cloud sessions.
Tool execution: Supports
Tool execution, supports. Agents run terminal commands and call MCP tools under allow, ask, and deny rules that admins can enforce; tool orchestration serves the coding work.
Validate
Debugging: Supports
Debugging, supports. Developers can send Problems panel errors and terminal output straight to the agent, which investigates them with repository context.
Testing: Supports
Testing, supports. Agents can write and run tests in the editor's terminal, but Devin Desktop does not replace dedicated testing infrastructure.
Code review: Supports
Code review, supports. Quick Review runs a second agent over local Devin Local changes before commit. Pull request review is a separate Cognition product, Devin Review.
Application security: Touches
Application security, touches. Enterprise admins can enforce an OS-level sandbox and domain rules for local agents, but Devin Desktop is not an application-security platform.
Ship
Source control: Touches
Source control, touches. Agents work in git worktrees and the editor drafts commit messages; Git stays the system of record.
CI/CD: Touches
CI/CD, touches. Devin Desktop has no app deploy since Cascade was removed on Sep 8, 2026; delivery stays in the team's CI/CD.
Operate
Observability: Touches
Observability, touches. Agents can help investigate production issues when given logs or traces, but Devin Desktop provides no observability itself.
Used for feature development, code generation, refactoring, debugging, codebase exploration, and testing, by software engineers, engineering teams, and platform teams.
- Business Area
- Product & Engineering
- Function
- Software Development
- Operating Stack
- Agentic Software Development
- Primary Role
- Agent Editor & Agent Command Center
How to read this map. Each layer lists the products that structurally own it, so a product that only touches a layer isn't listed among its owners. Devin Desktop (formerly Windsurf) is underlined in teal where it is Core, and its rows sit a shade lighter; where it supports or touches a layer, that row names it first.
Plan
Devin Desktop (formerly Windsurf): supports on 1 of 1 layer
Work planning & task decomposition
Turning engineering goals, issues, and requirements into actionable implementation work.
- Devin Desktop (formerly Windsurf) supportsSupportingDevin Local's Plan mode researches the codebase and writes an editable plan before implementing, and Spaces group sessions by task; planning stays in the team's tracker.
- LinearTeam WorkspaceLinearLinear+41 PositiveSoftware issue tracking and planning with Triage, cycles, projects, and initiatives. GitHub and GitLab events update status, and teams can assign issues to Linear Agent or third-party coding agents.Owns this layerWork planning & task decompositionOpen product page
- Jira
- GitHub Issues
Build
Devin Desktop (formerly Windsurf): core on 3 of 4 layers; supports on 1 of 4 layers
Codebase context & intelligence
Understanding repository structure, relevant files, dependencies, conventions, and existing implementation context.
- Devin Desktop (formerly Windsurf)
- AiderAI Coding AgentAiderAider AI LLC+28 PositiveOpen-source command-line pair programmer, licensed Apache 2.0, that edits the files you add to the chat with help from a map of the whole repository. It commits each change to git and connects to almost any LLM through your own key, including local models.Owns this layerCodebase context & intelligenceOpen product page
- Auggie CLIAI Coding AgentAuggie CLIHarness+59 Strong positiveCoding agent for the terminal and CI, built by Augment Code and owned by Harness since October 8, 2026, when Harness acquired Auggie CLI, Cosmos, and the Context Engine. It shares its agent and Context Engine with the Augment extensions for VS Code and JetBrains IDEs.Owns this layerCodebase context & intelligenceOpen product page
- Claude CodeAI Coding AgentClaude CodeAnthropic+75 Strong positiveReads a repository, edits files, runs commands, and returns changes for review, from the terminal, IDEs, a desktop app, the web, and CI.Owns this layerCodebase context & intelligenceOpen product page
- CodexAI Coding AgentCodexOpenAI+55 Strong positiveReads, edits, and runs code in a sandbox, locally through the CLI, IDE extension, and ChatGPT desktop app, or in hosted Codex Cloud environments.Owns this layerCodebase context & intelligenceOpen product page
- CursorAI Coding AgentCursorAnysphere+11 MixedAI coding agent and code editor for writing, reviewing, and shipping software, with agents in the editor, the terminal, and the cloud.Owns this layerCodebase context & intelligenceOpen product page
- JunieAI Coding AgentJunieJetBrains+28 PositiveAI coding agent from JetBrains that plans and makes multi-step code changes inside JetBrains IDEs, using their index, test runners, and debugger. It also runs from Junie CLI in the terminal, GitHub Actions, and GitLab CI/CD, on models from several AI labs or your own keys.Owns this layerCodebase context & intelligenceOpen product page
- kcapAI ObservabilitykcapKurrentUnscored until 20 sources from 15 authorskcap, formerly Kurrent Capacitor, records coding agents' sessions into a team workspace as they run: prompts, replies, every tool call and its result, and the commits made. Teammates search and review them, and the next agent reads them back through kcap's MCP servers. Team is priced per seat, a local edition is free for one user, and access is by invite.Owns this layerCodebase context & intelligenceOpen product page
- TabnineAI Coding AgentTabnineTabnine (a Tricentis company)−34 NegativeTabnine provides code completions, chat, an IDE agent, and a CLI inside VS Code, JetBrains IDEs, Eclipse, Visual Studio, and Cursor, on models the organization chooses. It adds a Context Engine over connected repositories and provenance checks against public code. Tricentis acquired Tabnine in July 2026.Owns this layerCodebase context & intelligenceOpen product page
- Sourcegraph
Coding & implementation
Creating and modifying production code to complete engineering work.
- Devin Desktop (formerly Windsurf)
- AiderAI Coding AgentAiderAider AI LLC+28 PositiveOpen-source command-line pair programmer, licensed Apache 2.0, that edits the files you add to the chat with help from a map of the whole repository. It commits each change to git and connects to almost any LLM through your own key, including local models.Owns this layerCoding & implementationOpen product page
- Amazon Q DeveloperAI Coding AgentAmazon Q DeveloperAmazon Web Services+10 MixedAWS's generative AI assistant for building and operating on AWS. It offers inline suggestions and agentic chat in VS Code, JetBrains IDEs, Visual Studio, and Eclipse. Amazon Q also works in the AWS console, Slack, and Microsoft Teams. AWS's end-of-support notice closed new sign-ups on May 15, 2026, though its pricing page still offers them. Support for the IDE plugins and paid subscriptions ends April 30, 2027, and AWS points coding work to Kiro.Owns this layerCoding & implementationOpen product page
- AmpAI Coding AgentAmpAmp Frontier Corporation+69 Strong positiveAmp is a coding agent from Amp Frontier Corporation, spun out of Sourcegraph in December 2025. Engineers use the CLI, web, macOS, iOS, or Android apps. Threads run locally, on their own runner, or in an orb, a cloud machine Amp creates per thread. Amp chooses Anthropic, OpenAI, and other models per mode.Owns this layerCoding & implementationOpen product page
- Google AntigravityAI Coding AgentGoogle AntigravityGoogle−16 MixedGoogle's agent platform for software work. Agents plan, write code, run terminal commands, and test in a browser. They run from the Antigravity 2.0 desktop app, the CLI, the original IDE, or extensions for VS Code, Visual Studio, JetBrains, Zed, and Xcode. It runs Gemini models and, on eligible plans, Anthropic's Claude models.Owns this layerCoding & implementationOpen product page
- Auggie CLIAI Coding AgentAuggie CLIHarness+59 Strong positiveCoding agent for the terminal and CI, built by Augment Code and owned by Harness since October 8, 2026, when Harness acquired Auggie CLI, Cosmos, and the Context Engine. It shares its agent and Context Engine with the Augment extensions for VS Code and JetBrains IDEs.Owns this layerCoding & implementationOpen product page
- BoltAI App BuilderBoltStackBlitz+60 Strong positiveStackBlitz's AI app builder at bolt.new. Its agent writes web apps in any JavaScript framework, or Expo mobile apps, inside a development environment that runs in the browser on WebContainers. Bolt Cloud adds a database, sign-in, and storage, and publishes to bolt.host. Projects sync to GitHub, and usage is metered in tokens.Owns this layerCoding & implementationOpen product page
- Claude CodeAI Coding AgentClaude CodeAnthropic+75 Strong positiveReads a repository, edits files, runs commands, and returns changes for review, from the terminal, IDEs, a desktop app, the web, and CI.Owns this layerCoding & implementationOpen product page
- ClineAI Coding AgentClineCline+39 PositiveOpen-source coding agent that plans changes, then edits files and runs terminal commands with your approval. It runs from VS Code, JetBrains IDEs, a CLI, or a desktop app, on the model provider you choose.Owns this layerCoding & implementationOpen product page
- CodexAI Coding AgentCodexOpenAI+55 Strong positiveReads, edits, and runs code in a sandbox, locally through the CLI, IDE extension, and ChatGPT desktop app, or in hosted Codex Cloud environments.Owns this layerCoding & implementationOpen product page
- ContinueAI Coding AgentContinueContinue Dev, Inc. (acquired by Cursor)−9 MixedOpen-source coding assistant and agent for VS Code, JetBrains IDEs, and the terminal, with chat, edits, tab autocomplete, and Agent mode on the model provider or local models you configure. Cursor acquired Continue in June 2026; the hosted service closed and the Apache 2.0 repository is read-only.Owns this layerCoding & implementationOpen product page
- CursorAI Coding AgentCursorAnysphere+11 MixedAI coding agent and code editor for writing, reviewing, and shipping software, with agents in the editor, the terminal, and the cloud.Owns this layerCoding & implementationOpen product page
- DevinAI Coding AgentDevinCognition+39 PositiveCognition's autonomous coding agent. It takes a ticket or a Slack thread, works in its own cloud machine, and returns a pull request. A terminal CLI and pull request review come with it.Owns this layerCoding & implementationOpen product page
- FactoryAI Coding AgentFactoryFactory+22 PositiveFactory's agent, Droid, plans, writes, tests, and reviews code from a terminal CLI, a desktop and web app, IDEs, CI jobs, and Slack or Linear delegations. It runs models from Anthropic, OpenAI, Google, xAI, and open-weight providers.Owns this layerCoding & implementationOpen product page
- Gemini CLIAI Coding AgentGemini CLIGoogle−3 MixedGoogle's open-source terminal agent edits code and runs shell commands and MCP tools. It reaches Gemini models through an existing Gemini Code Assist subscription, the Gemini API, or Vertex AI.Owns this layerCoding & implementationOpen product page
- GitHub CopilotAI Coding AgentGitHub CopilotGitHub+15 MixedAI coding product from GitHub with completions and chat in the IDE, agents in the IDE, the CLI, a desktop app, and the cloud, and code review on GitHub pull requests.Owns this layerCoding & implementationOpen product page
- JulesAI Coding AgentJulesGoogle0 MixedGoogle Labs' asynchronous coding agent. It clones a GitHub repository into a cloud VM, proposes a plan, writes and tests the change, and returns a branch or pull request. Task limits rise with Google AI Pro and Ultra plans on personal Google accounts.Owns this layerCoding & implementationOpen product page
- JunieAI Coding AgentJunieJetBrains+28 PositiveAI coding agent from JetBrains that plans and makes multi-step code changes inside JetBrains IDEs, using their index, test runners, and debugger. It also runs from Junie CLI in the terminal, GitHub Actions, and GitLab CI/CD, on models from several AI labs or your own keys.Owns this layerCoding & implementationOpen product page
- Kimi CodeAI Coding AgentKimi CodeMoonshot AI−36 NegativeKimi Code is Moonshot AI's coding agent for the terminal, VS Code, the desktop, and Zed or JetBrains over the Agent Client Protocol, running Kimi models up to K3.1 Pro with a 1M context window. It comes with a Kimi subscription from Plus at $19 a month (Kimi's plan page shows $ without naming a currency), all clients share one quota, and Extra Usage is charged by actual usage once the quota runs out.Owns this layerCoding & implementationOpen product page
- KiroAI Coding AgentKiroAmazon Web Services+10 MixedAgentic IDE and CLI from AWS that turns prompts into requirements, designs, and task lists before writing code. It adds cloud sessions, steering files, hooks, and a choice of Claude, GPT, and open-weight models billed in credits.Owns this layerCoding & implementationOpen product page
- LovableAI App BuilderLovableLovable+27 PositiveAI app builder that writes full-stack web apps from chat, with a built-in Supabase-based backend, Lovable hosting, and two-way Git sync, priced by workspace credits.Owns this layerCoding & implementationOpen product page
- Mistral VibeAI Coding AgentMistral VibeMistral AI+18 MixedMistral Vibe is Mistral AI's coding agent. Its Apache 2.0 CLI reads a repository, edits files, and runs shell commands and MCP tools under per-tool approvals, on Mistral-hosted models through a Vibe plan or on a local model server. A VS Code extension and Vibe Code Web run the same agent.Owns this layerCoding & implementationOpen product page
- OpenCodeAI Coding AgentOpenCodeAnomaly+27 PositiveOpen-source, MIT-licensed coding agent that edits code and runs commands from a terminal UI, desktop app, web client, or editor extension. It runs on models from 75+ vendors, local models, or an existing GitHub Copilot or ChatGPT plan.Owns this layerCoding & implementationOpen product page
- OpenHandsAI Coding AgentOpenHandsAll Hands AI+15 MixedAll Hands AI's MIT-licensed coding agent, formerly OpenDevin, writes and runs code in a sandbox and opens pull requests using the model you connect. It runs locally, on your servers, in OpenHands Cloud, or as self-hosted OpenHands Enterprise.Owns this layerCoding & implementationOpen product page
- Qwen CodeAI Coding AgentQwen CodeAlibaba−50 Strong negativeApache 2.0 coding agent from Alibaba's Qwen team that edits code and runs commands from the terminal, IDE plugins, a desktop app, and a web UI. It began as a fork of Gemini CLI and calls OpenAI, Anthropic, Gemini, and Qwen APIs or local models through Ollama or vLLM. The free Qwen OAuth tier ended on April 15, 2026.Owns this layerCoding & implementationOpen product page
- Replit AgentAI App BuilderReplit AgentReplit+26 PositiveAI app-building agent from Replit that plans, writes, tests, and publishes full-stack apps inside Replit's hosted workspace, with built-in PostgreSQL databases, authentication, secrets, and deployments.Owns this layerCoding & implementationOpen product page
- TabnineAI Coding AgentTabnineTabnine (a Tricentis company)−34 NegativeTabnine provides code completions, chat, an IDE agent, and a CLI inside VS Code, JetBrains IDEs, Eclipse, Visual Studio, and Cursor, on models the organization chooses. It adds a Context Engine over connected repositories and provenance checks against public code. Tricentis acquired Tabnine in July 2026.Owns this layerCoding & implementationOpen product page
- v0AI App Builderv0Vercel+68 Strong positiveAI app builder from Vercel that generates React and Next.js interfaces and full-stack apps from prompts, screenshots, and Figma designs. It runs them in an isolated sandbox, and publishes to Vercel or through GitHub pull requests.Owns this layerCoding & implementationOpen product page
- WarpAI Coding AgentWarpWarp+30 PositiveWarp is a terminal for macOS, Linux, and Windows with a built-in coding agent. The Warp Agent also runs as a CLI in any terminal and as cloud agents triggered from Slack, Linear, GitHub, or schedules. It uses models from OpenAI, Anthropic, Google, xAI, and open-weight providers. The client is open source under AGPL v3.Owns this layerCoding & implementationOpen product page
- ZedAI Coding AgentZedZed Industries+25 PositiveZed is an open-source code editor written in Rust for macOS, Windows, and Linux. It runs the Zed Agent and external agents such as Claude Agent, Codex, and Gemini CLI through the Agent Client Protocol, predicts edits with its Zeta model, and supports real-time multiplayer editing.Owns this layerCoding & implementationOpen product page
Agent interface & developer runtime
The environment through which a developer delegates, steers, and interacts with an engineering agent.
- Devin Desktop (formerly Windsurf)
- AiderAI Coding AgentAiderAider AI LLC+28 PositiveOpen-source command-line pair programmer, licensed Apache 2.0, that edits the files you add to the chat with help from a map of the whole repository. It commits each change to git and connects to almost any LLM through your own key, including local models.Owns this layerAgent interface & developer runtimeOpen product page
- Amazon Q DeveloperAI Coding AgentAmazon Q DeveloperAmazon Web Services+10 MixedAWS's generative AI assistant for building and operating on AWS. It offers inline suggestions and agentic chat in VS Code, JetBrains IDEs, Visual Studio, and Eclipse. Amazon Q also works in the AWS console, Slack, and Microsoft Teams. AWS's end-of-support notice closed new sign-ups on May 15, 2026, though its pricing page still offers them. Support for the IDE plugins and paid subscriptions ends April 30, 2027, and AWS points coding work to Kiro.Owns this layerAgent interface & developer runtimeOpen product page
- AmpAI Coding AgentAmpAmp Frontier Corporation+69 Strong positiveAmp is a coding agent from Amp Frontier Corporation, spun out of Sourcegraph in December 2025. Engineers use the CLI, web, macOS, iOS, or Android apps. Threads run locally, on their own runner, or in an orb, a cloud machine Amp creates per thread. Amp chooses Anthropic, OpenAI, and other models per mode.Owns this layerAgent interface & developer runtimeOpen product page
- Google AntigravityAI Coding AgentGoogle AntigravityGoogle−16 MixedGoogle's agent platform for software work. Agents plan, write code, run terminal commands, and test in a browser. They run from the Antigravity 2.0 desktop app, the CLI, the original IDE, or extensions for VS Code, Visual Studio, JetBrains, Zed, and Xcode. It runs Gemini models and, on eligible plans, Anthropic's Claude models.Owns this layerAgent interface & developer runtimeOpen product page
- Auggie CLIAI Coding AgentAuggie CLIHarness+59 Strong positiveCoding agent for the terminal and CI, built by Augment Code and owned by Harness since October 8, 2026, when Harness acquired Auggie CLI, Cosmos, and the Context Engine. It shares its agent and Context Engine with the Augment extensions for VS Code and JetBrains IDEs.Owns this layerAgent interface & developer runtimeOpen product page
- BoltAI App BuilderBoltStackBlitz+60 Strong positiveStackBlitz's AI app builder at bolt.new. Its agent writes web apps in any JavaScript framework, or Expo mobile apps, inside a development environment that runs in the browser on WebContainers. Bolt Cloud adds a database, sign-in, and storage, and publishes to bolt.host. Projects sync to GitHub, and usage is metered in tokens.Owns this layerAgent interface & developer runtimeOpen product page
- Claude CodeAI Coding AgentClaude CodeAnthropic+75 Strong positiveReads a repository, edits files, runs commands, and returns changes for review, from the terminal, IDEs, a desktop app, the web, and CI.Owns this layerAgent interface & developer runtimeOpen product page
- ClineAI Coding AgentClineCline+39 PositiveOpen-source coding agent that plans changes, then edits files and runs terminal commands with your approval. It runs from VS Code, JetBrains IDEs, a CLI, or a desktop app, on the model provider you choose.Owns this layerAgent interface & developer runtimeOpen product page
- CodexAI Coding AgentCodexOpenAI+55 Strong positiveReads, edits, and runs code in a sandbox, locally through the CLI, IDE extension, and ChatGPT desktop app, or in hosted Codex Cloud environments.Owns this layerAgent interface & developer runtimeOpen product page
- ContinueAI Coding AgentContinueContinue Dev, Inc. (acquired by Cursor)−9 MixedOpen-source coding assistant and agent for VS Code, JetBrains IDEs, and the terminal, with chat, edits, tab autocomplete, and Agent mode on the model provider or local models you configure. Cursor acquired Continue in June 2026; the hosted service closed and the Apache 2.0 repository is read-only.Owns this layerAgent interface & developer runtimeOpen product page
- CursorAI Coding AgentCursorAnysphere+11 MixedAI coding agent and code editor for writing, reviewing, and shipping software, with agents in the editor, the terminal, and the cloud.Owns this layerAgent interface & developer runtimeOpen product page
- FactoryAI Coding AgentFactoryFactory+22 PositiveFactory's agent, Droid, plans, writes, tests, and reviews code from a terminal CLI, a desktop and web app, IDEs, CI jobs, and Slack or Linear delegations. It runs models from Anthropic, OpenAI, Google, xAI, and open-weight providers.Owns this layerAgent interface & developer runtimeOpen product page
- Gemini CLIAI Coding AgentGemini CLIGoogle−3 MixedGoogle's open-source terminal agent edits code and runs shell commands and MCP tools. It reaches Gemini models through an existing Gemini Code Assist subscription, the Gemini API, or Vertex AI.Owns this layerAgent interface & developer runtimeOpen product page
- GitHub CopilotAI Coding AgentGitHub CopilotGitHub+15 MixedAI coding product from GitHub with completions and chat in the IDE, agents in the IDE, the CLI, a desktop app, and the cloud, and code review on GitHub pull requests.Owns this layerAgent interface & developer runtimeOpen product page
- JunieAI Coding AgentJunieJetBrains+28 PositiveAI coding agent from JetBrains that plans and makes multi-step code changes inside JetBrains IDEs, using their index, test runners, and debugger. It also runs from Junie CLI in the terminal, GitHub Actions, and GitLab CI/CD, on models from several AI labs or your own keys.Owns this layerAgent interface & developer runtimeOpen product page
- Kimi CodeAI Coding AgentKimi CodeMoonshot AI−36 NegativeKimi Code is Moonshot AI's coding agent for the terminal, VS Code, the desktop, and Zed or JetBrains over the Agent Client Protocol, running Kimi models up to K3.1 Pro with a 1M context window. It comes with a Kimi subscription from Plus at $19 a month (Kimi's plan page shows $ without naming a currency), all clients share one quota, and Extra Usage is charged by actual usage once the quota runs out.Owns this layerAgent interface & developer runtimeOpen product page
- KiroAI Coding AgentKiroAmazon Web Services+10 MixedAgentic IDE and CLI from AWS that turns prompts into requirements, designs, and task lists before writing code. It adds cloud sessions, steering files, hooks, and a choice of Claude, GPT, and open-weight models billed in credits.Owns this layerAgent interface & developer runtimeOpen product page
- LovableAI App BuilderLovableLovable+27 PositiveAI app builder that writes full-stack web apps from chat, with a built-in Supabase-based backend, Lovable hosting, and two-way Git sync, priced by workspace credits.Owns this layerAgent interface & developer runtimeOpen product page
- Mistral VibeAI Coding AgentMistral VibeMistral AI+18 MixedMistral Vibe is Mistral AI's coding agent. Its Apache 2.0 CLI reads a repository, edits files, and runs shell commands and MCP tools under per-tool approvals, on Mistral-hosted models through a Vibe plan or on a local model server. A VS Code extension and Vibe Code Web run the same agent.Owns this layerAgent interface & developer runtimeOpen product page
- OpenCodeAI Coding AgentOpenCodeAnomaly+27 PositiveOpen-source, MIT-licensed coding agent that edits code and runs commands from a terminal UI, desktop app, web client, or editor extension. It runs on models from 75+ vendors, local models, or an existing GitHub Copilot or ChatGPT plan.Owns this layerAgent interface & developer runtimeOpen product page
- OpenHandsAI Coding AgentOpenHandsAll Hands AI+15 MixedAll Hands AI's MIT-licensed coding agent, formerly OpenDevin, writes and runs code in a sandbox and opens pull requests using the model you connect. It runs locally, on your servers, in OpenHands Cloud, or as self-hosted OpenHands Enterprise.Owns this layerAgent interface & developer runtimeOpen product page
- Qwen CodeAI Coding AgentQwen CodeAlibaba−50 Strong negativeApache 2.0 coding agent from Alibaba's Qwen team that edits code and runs commands from the terminal, IDE plugins, a desktop app, and a web UI. It began as a fork of Gemini CLI and calls OpenAI, Anthropic, Gemini, and Qwen APIs or local models through Ollama or vLLM. The free Qwen OAuth tier ended on April 15, 2026.Owns this layerAgent interface & developer runtimeOpen product page
- Replit AgentAI App BuilderReplit AgentReplit+26 PositiveAI app-building agent from Replit that plans, writes, tests, and publishes full-stack apps inside Replit's hosted workspace, with built-in PostgreSQL databases, authentication, secrets, and deployments.Owns this layerAgent interface & developer runtimeOpen product page
- T3 CodeAI Coding AgentT3 CodeT3 Tools+20 PositiveMIT-licensed app from T3 Tools that runs Codex, Claude Code, Cursor, Grok Build, OpenCode, Antigravity and fx through the subscriptions and logins you already have, from a desktop app, a local web app or a phone.Owns this layerAgent interface & developer runtimeOpen product page
- v0AI App Builderv0Vercel+68 Strong positiveAI app builder from Vercel that generates React and Next.js interfaces and full-stack apps from prompts, screenshots, and Figma designs. It runs them in an isolated sandbox, and publishes to Vercel or through GitHub pull requests.Owns this layerAgent interface & developer runtimeOpen product page
- WarpAI Coding AgentWarpWarp+30 PositiveWarp is a terminal for macOS, Linux, and Windows with a built-in coding agent. The Warp Agent also runs as a CLI in any terminal and as cloud agents triggered from Slack, Linear, GitHub, or schedules. It uses models from OpenAI, Anthropic, Google, xAI, and open-weight providers. The client is open source under AGPL v3.Owns this layerAgent interface & developer runtimeOpen product page
- ZedAI Coding AgentZedZed Industries+25 PositiveZed is an open-source code editor written in Rust for macOS, Windows, and Linux. It runs the Zed Agent and external agents such as Claude Agent, Codex, and Gemini CLI through the Agent Client Protocol, predicts edits with its Zeta model, and supports real-time multiplayer editing.Owns this layerAgent interface & developer runtimeOpen product page
Tool execution & integrations
Allowing the agent to invoke development tools, commands, connected systems, and authorized external capabilities.
- Devin Desktop (formerly Windsurf) supportsSupportingAgents run terminal commands and call MCP tools under allow, ask, and deny rules that admins can enforce; tool orchestration serves the coding work.
- Google AntigravityAI Coding AgentGoogle AntigravityGoogle−16 MixedGoogle's agent platform for software work. Agents plan, write code, run terminal commands, and test in a browser. They run from the Antigravity 2.0 desktop app, the CLI, the original IDE, or extensions for VS Code, Visual Studio, JetBrains, Zed, and Xcode. It runs Gemini models and, on eligible plans, Anthropic's Claude models.Owns this layerTool execution & integrationsOpen product page
- Claude CodeAI Coding AgentClaude CodeAnthropic+75 Strong positiveReads a repository, edits files, runs commands, and returns changes for review, from the terminal, IDEs, a desktop app, the web, and CI.Owns this layerTool execution & integrationsOpen product page
- ClineAI Coding AgentClineCline+39 PositiveOpen-source coding agent that plans changes, then edits files and runs terminal commands with your approval. It runs from VS Code, JetBrains IDEs, a CLI, or a desktop app, on the model provider you choose.Owns this layerTool execution & integrationsOpen product page
- CodexAI Coding AgentCodexOpenAI+55 Strong positiveReads, edits, and runs code in a sandbox, locally through the CLI, IDE extension, and ChatGPT desktop app, or in hosted Codex Cloud environments.Owns this layerTool execution & integrationsOpen product page
- DevinAI Coding AgentDevinCognition+39 PositiveCognition's autonomous coding agent. It takes a ticket or a Slack thread, works in its own cloud machine, and returns a pull request. A terminal CLI and pull request review come with it.Owns this layerTool execution & integrationsOpen product page
- FactoryAI Coding AgentFactoryFactory+22 PositiveFactory's agent, Droid, plans, writes, tests, and reviews code from a terminal CLI, a desktop and web app, IDEs, CI jobs, and Slack or Linear delegations. It runs models from Anthropic, OpenAI, Google, xAI, and open-weight providers.Owns this layerTool execution & integrationsOpen product page
- Gemini CLIAI Coding AgentGemini CLIGoogle−3 MixedGoogle's open-source terminal agent edits code and runs shell commands and MCP tools. It reaches Gemini models through an existing Gemini Code Assist subscription, the Gemini API, or Vertex AI.Owns this layerTool execution & integrationsOpen product page
- JulesAI Coding AgentJulesGoogle0 MixedGoogle Labs' asynchronous coding agent. It clones a GitHub repository into a cloud VM, proposes a plan, writes and tests the change, and returns a branch or pull request. Task limits rise with Google AI Pro and Ultra plans on personal Google accounts.Owns this layerTool execution & integrationsOpen product page
- Kimi CodeAI Coding AgentKimi CodeMoonshot AI−36 NegativeKimi Code is Moonshot AI's coding agent for the terminal, VS Code, the desktop, and Zed or JetBrains over the Agent Client Protocol, running Kimi models up to K3.1 Pro with a 1M context window. It comes with a Kimi subscription from Plus at $19 a month (Kimi's plan page shows $ without naming a currency), all clients share one quota, and Extra Usage is charged by actual usage once the quota runs out.Owns this layerTool execution & integrationsOpen product page
- Mistral VibeAI Coding AgentMistral VibeMistral AI+18 MixedMistral Vibe is Mistral AI's coding agent. Its Apache 2.0 CLI reads a repository, edits files, and runs shell commands and MCP tools under per-tool approvals, on Mistral-hosted models through a Vibe plan or on a local model server. A VS Code extension and Vibe Code Web run the same agent.Owns this layerTool execution & integrationsOpen product page
- OpenCodeAI Coding AgentOpenCodeAnomaly+27 PositiveOpen-source, MIT-licensed coding agent that edits code and runs commands from a terminal UI, desktop app, web client, or editor extension. It runs on models from 75+ vendors, local models, or an existing GitHub Copilot or ChatGPT plan.Owns this layerTool execution & integrationsOpen product page
- OpenHandsAI Coding AgentOpenHandsAll Hands AI+15 MixedAll Hands AI's MIT-licensed coding agent, formerly OpenDevin, writes and runs code in a sandbox and opens pull requests using the model you connect. It runs locally, on your servers, in OpenHands Cloud, or as self-hosted OpenHands Enterprise.Owns this layerTool execution & integrationsOpen product page
- Qwen CodeAI Coding AgentQwen CodeAlibaba−50 Strong negativeApache 2.0 coding agent from Alibaba's Qwen team that edits code and runs commands from the terminal, IDE plugins, a desktop app, and a web UI. It began as a fork of Gemini CLI and calls OpenAI, Anthropic, Gemini, and Qwen APIs or local models through Ollama or vLLM. The free Qwen OAuth tier ended on April 15, 2026.Owns this layerTool execution & integrationsOpen product page
- WarpAI Coding AgentWarpWarp+30 PositiveWarp is a terminal for macOS, Linux, and Windows with a built-in coding agent. The Warp Agent also runs as a CLI in any terminal and as cloud agents triggered from Slack, Linear, GitHub, or schedules. It uses models from OpenAI, Anthropic, Google, xAI, and open-weight providers. The client is open source under AGPL v3.Owns this layerTool execution & integrationsOpen product page
Validate
Devin Desktop (formerly Windsurf): supports on 3 of 4 layers; touches on 1 of 4 layers
Debugging & investigation
Diagnosing failures, understanding unexpected behavior, and locating likely causes.
- Devin Desktop (formerly Windsurf) supportsSupportingDevelopers can send Problems panel errors and terminal output straight to the agent, which investigates them with repository context.
- SeerAI ObservabilitySeerSentry+11 MixedSentry's AI debugging agent finds root causes in issue and trace data, along with logs and profiles. It drafts fixes or hands them to coding agents. It also reviews pull requests on GitHub.com and GitLab.com.Owns this layerDebugging & investigationOpen product page
Testing & verification
Creating, running, and interpreting checks that validate whether a change behaves as intended.
- Devin Desktop (formerly Windsurf) supportsSupportingAgents can write and run tests in the editor's terminal, but Devin Desktop does not replace dedicated testing infrastructure.
- Playwright
- Cypress
Code review & quality
Reviewing proposed changes for correctness, maintainability, architecture, and code quality.
- Devin Desktop (formerly Windsurf) supportsSupportingQuick Review runs a second agent over local Devin Local changes before commit. Pull request review is a separate Cognition product, Devin Review.
- CodeRabbitAI Code ReviewCodeRabbitCodeRabbit+16 MixedAI code review for pull requests, the IDE, and the CLI, with plans that add PR triage, test generation, and security review.Owns this layerCode review & qualityOpen product page
- GitHub CopilotAI Coding AgentGitHub CopilotGitHub+15 MixedAI coding product from GitHub with completions and chat in the IDE, agents in the IDE, the CLI, a desktop app, and the cloud, and code review on GitHub pull requests.Owns this layerCode review & qualityOpen product page
- GreptileAI Code ReviewGreptileGreptile+40 PositiveAI code review agent that reviews every pull request against a graph of the whole repository, in Greptile's cloud or self-hosted.Owns this layerCode review & qualityOpen product page
- QodoAI Code ReviewQodoQodo+17 MixedQodo reviews pull requests on GitHub, GitLab, Bitbucket, Azure DevOps, and Gerrit against your rules and repository context. Qodo Merge was the former name.Owns this layerCode review & qualityOpen product page
Application security
Finding, preventing, and managing security weaknesses in application code and dependencies.
- Devin Desktop (formerly Windsurf) touchesAdjacentEnterprise admins can enforce an OS-level sandbox and domain rules for local agents, but Devin Desktop is not an application-security platform.
- Snyk
- Semgrep
Ship
Devin Desktop (formerly Windsurf): touches on 2 of 2 layers
Source control & change management
Managing branches, commits, pull requests, merges, and the canonical history of code changes.
- Devin Desktop (formerly Windsurf) touchesAdjacentAgents work in git worktrees and the editor drafts commit messages; Git stays the system of record.
- GitHub
- GitLab
CI/CD & deployment
Building, validating, releasing, and deploying software through automated delivery systems.
- Devin Desktop (formerly Windsurf) touchesAdjacentDevin Desktop has no app deploy since Cascade was removed on Sep 8, 2026; delivery stays in the team's CI/CD.
- Replit AgentAI App BuilderReplit AgentReplit+26 PositiveAI app-building agent from Replit that plans, writes, tests, and publishes full-stack apps inside Replit's hosted workspace, with built-in PostgreSQL databases, authentication, secrets, and deployments.Owns this layerCI/CD & deploymentOpen product page
- GitHub Actions
- CircleCI
- Vercel
Operate
Devin Desktop (formerly Windsurf): touches on 1 of 1 layer
Observability & incident response
Monitoring production behavior, diagnosing incidents, and understanding system performance after release.
- Devin Desktop (formerly Windsurf) touchesAdjacentAgents can help investigate production issues when given logs or traces, but Devin Desktop provides no observability itself.
- SeerAI ObservabilitySeerSentry+11 MixedSentry's AI debugging agent finds root causes in issue and trace data, along with logs and profiles. It drafts fixes or hands them to coding agents. It also reviews pull requests on GitHub.com and GitLab.com.Owns this layerObservability & incident responseOpen product page
- Datadog
- Sentry
- PagerDuty