Tree source definition
open-source: Does the tool need to be open source?
Open source means you can audit the code, self-host, and avoid vendor lock-in — often a hard requirement for sensitive codebases.
os-form-factor: What form factor fits your workflow?
- A full standalone editor
- An extension inside my current IDE
- A self-hosted server for my team (air-gapped)
- A cockpit for running several coding agents in parallel
os-editor-style: What do you want from the editor?
- Speed and real-time collaboration, with AI built in
- A VS Code-like editor that talks straight to my own models, no middleman
- A framework to build my own IDE or cloud IDE
os-editor-privacy: How important is it that no vendor backend sees your code?
- Critical — prompts go only to the endpoint I configure
- Convenient model routing and agents matter more
extension-style: Which IDE do you live in, and what matters most?
- VS Code — autonomous agent workflows that edit files and run terminal commands
- VS Code or JetBrains — bring my own keys, run local models offline
- IntelliJ IDEA — local and cloud models with RAG over my project
agent-extension: Which agent style do you prefer?
- Plain, checkpointed human-in-the-loop control
- Customizable personas (modes) with configurable auto-approve
- Debugger modes, project memory and an MCP marketplace
selfhosted-style: What does your team need from a self-hosted assistant?
- Fast completion and chat running entirely on my own hardware
- An autonomous agent with repository-grounded retrieval (RAG)
os-orchestrate: What is the bottleneck?
- Juggling many parallel Claude Code sessions with worktrees
- Large repositories need compiler-grade context for the model
commercial-priority: What is your main priority?
- An AI-first editor with the strongest agents
- AI inside the IDE or ecosystem I already use
- Zero setup — work in the browser or a cloud workspace
- Enterprise: privacy, on-prem or very large codebases
switch-editor: Which style of AI-first editor do you want?
- Polished, with parallel agents and strong admin controls
- Cloud-hosted, understands the whole repository
- Spec-driven (requirements, design, tasks) and AWS-native
- Agent-first across editor, terminal and browser
- Multi-model with a planning-first Builder and an autonomous SOLO mode
stay-ecosystem: Which ecosystem do you already use?
- GitHub and Microsoft tools, any mainstream IDE
- JetBrains IDEs
- AWS
- Google Cloud — and a generous free tier matters
- Object Pascal
jetbrains-style: How autonomous should the JetBrains assistant be?
- An agent that validates its own output with IDE inspections and tests
- Chat, completion, refactoring and multi-file edits, with cloud or local models
cloud-workspace: What will you build in the cloud?
- Full-stack web or mobile prototypes with a Firebase backend
- Anything, in a browser IDE with an autonomous agent and built-in deploys
- Sandboxed team environments with agents and guardrails
enterprise-need: What does the enterprise need most?
- Strict privacy: private VPC, on-prem or air-gapped deployment
- Deep understanding of huge, multi-repo codebases
rec-zed: Zed
Zed — open-source (GPL-3.0 core) standalone editor. Supports OpenAI-compatible providers and local LLMs via Ollama, so you can bring your own keys or keep prompts and code entirely on-device. GPU-rendered and Rust-based, built for ultra-low-latency typing and real-time collaboration.
Citations
rec-void: Void
Void — open-source, privacy-first fork of Visual Studio Code. It connects directly to the local or cloud LLMs you configure and never acts as an intermediary for your prompts, which suits data-residency and on-prem constraints.
Citations
rec-pearai: PearAI
PearAI — open-source, AI-first, VS Code-like editor with BYOK, a model router, contextual @-commands, inline diffs and agentic automation (RooCode/Cline). The dataset flags early-stage rough edges and an initial licensing controversy, so check the current licence.
Citations
rec-theia: Eclipse Theia
Eclipse Theia — modular, vendor-neutral, open-source IDE framework for desktop and cloud, built with TypeScript and the Monaco editor. Compatible with VS Code extensions via Open VSX. Choose it when you want to embed or extend an IDE rather than adopt one.
Citations
rec-continue: Continue
Continue — open-source (Apache-2.0) assistant for VS Code and JetBrains, the terminal and CI/CD. Configure your own API keys and model endpoints, switch between local and remote models, and run local workflows via Ollama — including headless operation for air-gapped setups.
Citations
rec-devoxx: DevoxxGenie
DevoxxGenie — IntelliJ IDEA plugin with support for local and cloud models, RAG over your project, MCP integration and multimodal input.
Citations
rec-cline: Cline
Cline — open-source VS Code extension with agentic workflows: file editing, terminal and browser automation, MCP plugin support, and checkpointed human-in-the-loop control.
Citations
rec-roo: Roo Code
Roo Code — autonomous VS Code extension with customizable personas (modes), multi-provider BYOK support, terminal and browser automation via MCP, and configurable autonomy (auto-approve, checkpoints).
Citations
rec-kilo: Kilo Code
Kilo Code — open-source extension for VS Code and JetBrains with Architect, Coder and Debugger modes, project memory, a CLI and an MCP marketplace. Supports BYOK and local model runtimes.
Citations
rec-tabby: Tabby
Tabby — open-source (Apache 2.0), self-hosted AI coding assistant. Runs fully offline (Docker or from source) against local backends such as Ollama or llama.cpp — built for sensitive codebases and air-gapped environments.
Citations
rec-refact: Refact.ai
Refact.ai — open-source autonomous coding assistant with RAG-grounded completion, refactoring, in-IDE chat, test generation and autonomous repository operations. Docker images and docs are available for self-hosted and on-prem deployment.
Citations
rec-codelayer: CodeLayer
CodeLayer — open-source, keyboard-first IDE and orchestration layer built around Claude Code, with parallel sessions and worktree-aware git integration. It depends on Claude Code for its model, so fully local use is limited.
Citations
rec-brokk: Brokk
Brokk — open-source Java desktop application that gives models compiler-grade, workspace-driven context on large repositories (Deep Scan, Agentic Search, Code/Ask/Architect agents). It is GPL-3.0 (copyleft), which may matter for derivative work.
Citations
rec-cursor: Cursor
Cursor — commercial AI-first editor; the core platform and Composer model are proprietary. Treats agents as first-class collaborators: multi-agent parallel worktrees, an embedded browser and testing tools, and enterprise sandboxing and admin controls.
Citations
rec-windsurf: Windsurf
Windsurf — cloud-hosted agentic IDE that understands entire repositories and performs multi-file edits and iterative AI-driven development flows.
Citations
rec-kiro: Kiro
Kiro — AWS’s agentic IDE built on Code OSS. A spec-driven workflow (requirements, design, tasks) is the source of truth, with Agent Hooks on file save or PR open, and IDE, CLI and background-agent surfaces. It is locked to Amazon Bedrock models.
Citations
rec-antigravity: Google Antigravity
Google Antigravity — agentic IDE for delegating coding tasks across the editor, terminal and browser.
Citations
rec-trae: Trae
Trae — AI-first VS Code fork with multiple models, a planning-first Builder mode, the autonomous SOLO agent (requirements to deployment) and integrated deploys.
Citations
rec-copilot: GitHub Copilot
GitHub Copilot — AI pair-programmer from GitHub/Microsoft: inline completions, multi-file suggestions, Copilot Chat, and autonomous agent features, integrated across IDEs and github.com.
Citations
rec-junie: JetBrains Junie
Junie — coding agent inside every JetBrains IDE, with an autonomous Code Mode and a collaborative Ask Mode. After generating code it runs IDE inspections and your test suite to catch errors. A standalone Junie CLI supports BYOK.
Citations
rec-jb-assistant: JetBrains AI Assistant
JetBrains AI Assistant — project-aware completion, chat, refactoring, multi-file edits and test and documentation generation, with cloud or local models. Check whether your JetBrains subscription already bundles it.
Citations
rec-amazonq: Amazon Q Developer
Amazon Q Developer — AWS’s assistant for IDEs and the AWS Console: workspace-aware chat, inline actions, documentation and security scanning, MCP integrations. It is cloud-hosted with limited BYOK and no fully local mode.
Citations
rec-gemini: Gemini Code Assist
Gemini Code Assist — Google’s coding assistant with multimodal support, refactoring, deep Google Cloud integration, an agent mode and a large free tier (180K completions a month, per the dataset).
Citations
rec-codebot: CodeBot for Delphi
CodeBot for Delphi — Delphi-first coding agent inside the Delphi IDE that understands Object Pascal, VCL/FMX components and DFM form files.
Citations
rec-firebase: Firebase Studio
Firebase Studio — cloud IDE for prototyping and deploying full-stack web and mobile apps: Gemini conversational prototyping, a Code OSS workspace and one-click Firebase backend provisioning. Cloud-only and proprietary, so expect lock-in to Firebase and GCP.
Citations
rec-replit: Replit
Replit — browser-based cloud IDE with AI completion and the autonomous Replit Agent. Zero local setup, built-in deployment and collaboration.
Citations
rec-ona: Ona (formerly Gitpod)
Ona — the rebranded Gitpod: preconfigured, sandboxed cloud development environments with autonomous engineering agents and enterprise guardrails. Core Gitpod components stay open source; agent and enterprise features are commercial.
Citations
rec-private-deploy: Tabnine (or Mistral Code)
Tabnine — commercial assistant built around privacy and governance (SSO/SAML, SCIM, role-based access), with cloud, private-VPC, on-prem and air-gapped deployment and broad IDE support. Also consider Mistral Code, an enterprise assistant offering on-prem, cloud and air-gapped options.
Citations
rec-large-codebase: Augment Code (or Sourcegraph)
Augment Code — platform for deep codebase understanding with large-context (~200k tokens) completions, autonomous agents and multi-file Next Edit workflows; credit-based pricing. Also consider Sourcegraph for cross-repository code search and navigation (note its 2023–24 move away from Apache 2.0) and Jolt for multi-repo context.