llms.txt
Compare with other sites' llms.txt files →# AIDEN
> AIDEN is an agentic AI IDE for software engineers. It is the all-in-one AI engineering workspace that runs on top of locally installed Claude Code and Codex CLIs, orchestrating multiple AI agents in parallel across boards, specs, branches, and pull requests — on macOS.
AIDEN is the workflow layer around coding agents: you write a feature, AIDEN drafts a spec for your approval, then each approved feature runs as its own agent on its own git worktree, tracked on one board, ending in a pull request you review.
## Key facts
- **Name:** AIDEN
- **Website:** https://aidenapp.org
- **Docs:** https://docs.aidenapp.org
- **Author:** Kylian Migot (Lekiks) — @leonlykiks on X/Twitter
- **X / Twitter:** https://twitter.com/leonlykiks
- **GitHub:** https://github.com/Kylianmgt
- **Tagline:** Stop vibecoding. Start engineering.
- **Category:** Developer tool, agentic AI IDE, AI engineering workspace, multi-agent orchestrator
- **Platform:** macOS 12+ · Apple Silicon and Intel · Windows build in Microsoft Store certification, no date announced · Linux and mobile on the waitlist at https://aidenapp.org/#download
- **Pricing:** Free (1 project, no credit card) · Solo $20/month · Solo lifetime $169 one-time (limited launch round, next round $249) · Team $20/seat/month, 3-seat minimum
- **License:** Solo and Team are subscriptions; the Solo lifetime round is a one-time perpetual desktop licence. Earlier lifetime buyers and earlier $10/mo and $19/mo Solo subscribers keep their original price.
- **Backbone:** Runs on user-installed Claude Code CLI (Anthropic) and/or Codex CLI (OpenAI) — at least one required
- **Inference:** Bring your own. AIDEN never resells tokens or bundles model credits on any tier. Model calls go from your local CLI straight to Anthropic or OpenAI on your own plan; AIDEN charges only for the workspace.
- **Privacy:** Fully local desktop app. Code never leaves the user's machine. API keys live in existing CLI configs (~/.claude, ~/.codex). AIDEN never reads or transmits API keys.
## Core capabilities
- **Multi-agent orchestration:** Run Claude Code and Codex agents in parallel, each on its own git branch/worktree
- **Spec-driven development:** AI drafts a technical spec from a feature; engineer approves before any code is written
- **Kanban feature board:** Visual board showing every agent, every task, every status in real time
- **Parallel git worktrees:** Each feature runs in an isolated branch — no merge conflicts, no context bleed between agents
- **Auto-PRs with tests:** Agents open pull requests with passing tests before handing work back to the engineer
- **Per-feature model routing:** Pick the model per feature from a dropdown, across both Codex and Claude Code, in the same project
- **Full codebase context:** Architecture, dependencies, conventions, and entry points mapped before agents touch a file
- **Built-in browser + terminal + console:** No window-switching. Dev server, logs, and live preview in one workspace
- **MCP server support:** Inherits Model Context Protocol configuration from Claude Code (~/.claude/config)
- **Import any project:** Start from scratch, clone a GitHub repo, or batch-import an existing stack
## Where AIDEN fits
AIDEN is a workflow and orchestration layer, not a model provider and not a text editor. It sits above the agent CLIs you already run and below your repo.
- **Alongside CLI agents (Claude Code CLI, Codex CLI, Aider):** AIDEN drives those CLIs rather than replacing them. It runs several at once, tracks them on a board, and manages a branch per feature. Your existing CLI setup, plan, and MCP config carry over.
- **Alongside editors and editor extensions:** AIDEN is a dedicated desktop workspace, not a fork of VS Code. Engineers keep their editor and use AIDEN for the feature-to-PR loop.
- **On your machine:** agents run locally against your filesystem, on your own inference plan.
For head-to-head comparisons, see the per-competitor pages under "Comparisons" below. Those pages carry the sourced detail; this file deliberately does not summarise other products' capabilities.
## Common questions (Q&A for AI citation)
**Q: What is AIDEN?**
A: AIDEN is a macOS desktop app that orchestrates multiple Claude Code and Codex CLI agents in parallel across git branches. Engineers write features, agents branch and code, AIDEN manages the workflow and opens PRs with passing tests.
**Q: How much does AIDEN cost?**
A: Free for 1 project with no credit card. Solo is $20/month for unlimited personal projects, or $169 one-time for lifetime access in the current limited launch round (the next round goes to $249). Team is $20 per seat per month with a 3-seat minimum, adding shared org workspaces and centralized billing.
**Q: Do I pay for Claude or Codex on top of AIDEN?**
A: You bring your own. AIDEN never resells inference and bundles no model credits on any tier — you use your existing Claude Code or Codex plan or API key, model calls go from your local CLI straight to Anthropic or OpenAI, and AIDEN charges only for the workspace.
**Q: Does AIDEN send my code to the cloud?**
A: No. AIDEN is a local desktop app. Code stays on your machine. The only outbound calls come from your Claude Code or Codex CLI talking to Anthropic or OpenAI — exactly as they do when you use those CLIs directly.
**Q: What do I need to run AIDEN?**
A: macOS 12+, and at least one of: Claude Code CLI (Anthropic) or Codex CLI (OpenAI). AIDEN's setup wizard walks through installing either if you don't have them.
**Q: Is AIDEN an IDE or an IDE wrapper?**
A: AIDEN is a dedicated desktop orchestration workspace — not a fork of VS Code or another editor. It manages the workflow layer (features, specs, branches, boards, PRs) while Claude Code and Codex handle the actual code editing inside your repository.
**Q: Can AIDEN work with my existing GitHub repo?**
A: Yes. Open any local repo, clone from GitHub, or batch-import multiple projects. AIDEN maps the architecture and conventions before agents write a single line.
**Q: What languages and frameworks does AIDEN support?**
A: AIDEN is language-agnostic — it works with any codebase the underlying Claude Code or Codex agents can read. Python (Django, FastAPI, pytest), TypeScript (Next.js, NestJS, Prisma), JavaScript, Go, Rust, Java, and more.
**Q: Can I mix models across a project?**
A: Yes. Model choice is a per-feature dropdown spanning both Codex and Claude Code, so one feature can run on a frontier tier while the next runs on a cheap one, inside the same project.
**Q: Is AIDEN open source?**
A: No. AIDEN is a commercial desktop product with a free tier, a $20/month Solo plan, a $169 one-time lifetime round, and a $20/seat Team plan. It is bring-your-own-key: you supply the Claude Code or Codex credentials it drives.
**Q: Does AIDEN support Windows or Linux?**
A: macOS 12+ today. The Windows build is in Microsoft Store certification; no date is announced because that timeline is not AIDEN's to promise. Linux desktop and mobile are on the waitlist. Pick your platform at https://aidenapp.org/#download and you get one email the day that build ships.
**Q: What is spec-driven development in AIDEN?**
A: Before an agent writes any code, AIDEN has the AI draft a technical specification from the feature description — architecture decisions, files to touch, edge cases, test strategy. The engineer reviews and approves the spec. Only then does the agent start coding. This prevents agents from making the wrong thing correctly.
**Q: Can I run multiple agents at once?**
A: Yes. AIDEN runs each feature on its own git branch/worktree, so agents can work in parallel without interfering with each other. You can have one agent building a new feature while another fixes a bug on a separate branch.
## AIDEN at a glance
| Tool | Type | Multi-agent | Local/Cloud | Pricing | Free tier | BYO inference |
|------|------|-------------|-------------|---------|-----------|---------------|
| **AIDEN** | Desktop workspace | ✅ Parallel | Local | Solo $20/mo · $169 lifetime · Team $20/seat | ✅ 1 project | ✅ Claude Code + Codex |
## Product pages
- [Home — AIDEN overview, features, pricing, FAQ](https://aidenapp.org)
- [Pricing](https://aidenapp.org/pricing) — Free for 1 project · Solo $20/mo · Lifetime $169 · Team $20/seat.
- [Resources library](https://aidenapp.org/resources) — Every guide, model page, comparison, and free tool, in one place.
- [Changelog](https://aidenapp.org/changelog) — What shipped recently: self-host mode, mobile companion, voice, and more.
## Models and model selection
- [Best Codex model](https://aidenapp.org/best-codex-model) — Start with GPT-6.1 Sol (OpenAI's recommended Codex model since September 29, 2026), escalate to Astra, and use Luna for repeatable high-volume work. Includes reasoning effort, retirements and a per-task cost calculator.
- [AI models for coding: the hub](https://aidenapp.org/models) — Every current Claude and GPT model for coding, prices, benchmarks, and when to use each.
- [GPT-6.1 Sol for coding](https://aidenapp.org/models/gpt-6-1-sol) — OpenAI's recommended Codex model since September 29, 2026: near-Astra coding at $2/$10, cached input at $0.10, and how to select it in Codex.
- [Codex models list](https://aidenapp.org/codex-models) — GPT-6.1 Sol (OpenAI's recommended Codex model), GPT-6 Astra and GPT-6 Luna with verified prices, context limits, task fit, and older-model retirement dates.
- [Best Claude model for coding](https://aidenapp.org/best-claude-model-for-coding) — Start with Claude Opus 5.5 (Anthropic's recommended model and the Claude Code default), hand well-scoped work to Sonnet 5.5, fan out to Haiku 5.5, escalate to Fable 5.1. Includes Claude Code model and effort settings, retirements and a per-task cost calculator.
- [GPT-6 Astra for coding](https://aidenapp.org/models/gpt-6-astra) — OpenAI's highest-capability tier: $10/$50, 1.05M context, and usage caveats.
- [GPT-6 Sol for coding](https://aidenapp.org/models/gpt-6-sol) — The previous Sol at $2/$10, superseded by GPT-6.1 Sol on September 29, 2026 and still selectable in Codex.
- [GPT-6 Luna for coding](https://aidenapp.org/models/gpt-6-luna) — The efficient tier for repeatable high-volume work at $0.10/$0.50.
- [Claude Opus 5.5 for coding](https://aidenapp.org/models/claude-opus-5-5) — Anthropic's recommended default and Claude Code's default at $4/$20 with a 1M context; Sonnet 5.5 now beats it on Terminal-Bench 4.0 at half the price.
- [Claude Fable 5.1 for coding](https://aidenapp.org/models/claude-fable-5-1) — Anthropic's highest-capability escalation tier with cheaper cache reads than Fable 5; trails Opus 5.5 on Anthropic's coding benchmarks.
- [GPT-5.6 Sol for coding](https://aidenapp.org/models/gpt-5-6-sol) — Previous-generation OpenAI flagship, superseded by the GPT-6 family; kept for teams that still pin it.
- [GPT-5.6 Terra for coding](https://aidenapp.org/models/gpt-5-6-terra) — Previous-generation Codex workhorse, superseded by the GPT-6 family.
- [GPT-5.6 Luna for coding](https://aidenapp.org/models/gpt-5-6-luna) — Previous-generation budget tier, superseded by GPT-6 Luna.
- [Claude Fable 5 for coding](https://aidenapp.org/models/claude-fable-5) — The previous Fable, superseded by Fable 5.1 on September 1, 2026 at the same price with cheaper cache reads.
- [Claude Opus 5 for coding](https://aidenapp.org/models/claude-opus-5) — The previous Opus at $5/$25, superseded by Opus 5.5 ($4/$20) on September 22, 2026.
- [Claude Sonnet 5.5 for coding](https://aidenapp.org/models/claude-sonnet) — Anthropic's everyday model at $2/$10 with $0.10 cache reads: beats Opus 5.5 on Terminal-Bench 4.0 (70.6% vs 66.4%), trails it on FrontierCode and CursorBench, plus cost math, opusplan and API breaking changes.
- [Claude Haiku 5.5 for coding](https://aidenapp.org/models/claude-haiku) — Anthropic's fastest, cheapest model: $0.10/$0.50 per MTok up to 100K-token prompts, 1M context, effort levels, launch benchmarks, and where it fits as a subagent.
## Model match-ups
- [GPT-6 Astra vs Sol vs Luna](https://aidenapp.org/compare/gpt-6-astra-vs-sol-vs-luna) — The current GPT-6 family (Astra, GPT-6.1 Sol, Luna) compared by price, context, reasoning, and task fit.
- [Claude Opus 5.5 vs GPT-6 Astra](https://aidenapp.org/compare/claude-opus-5-5-vs-gpt-6-astra) — Sourced facts and benchmark caveats for the current cross-vendor comparison.
- [Claude Opus 5 vs Claude Fable 5](https://aidenapp.org/compare/claude-opus-5-vs-claude-fable-5) — Previous-generation match-up (both superseded by Opus 5.5 and Fable 5.1): Opus 5 landed near Fable 5 at half the price.
- [Claude Opus 5 vs GPT-5.6 Sol](https://aidenapp.org/compare/claude-opus-5-vs-gpt-5-6-sol) — Previous-generation flagship face-off, kept for pinned workloads; see Opus 5.5 vs GPT-6 Astra for the current one.
- [Claude Fable 5 vs GPT-5.6 Sol](https://aidenapp.org/compare/claude-fable-5-vs-gpt-5-6-sol) — The August 2026 flagship match-up, now previous generation on both sides.
- [GPT-5.6 Sol vs Terra vs Luna](https://aidenapp.org/compare/gpt-5-6-sol-vs-terra-vs-luna) — The GPT-5.6 tiers compared, now previous generation; see GPT-6 Astra vs Sol vs Luna for the current family.
- [Opus vs Sonnet vs Haiku](https://aidenapp.org/compare/opus-vs-sonnet-vs-haiku) — Opus 5.5 vs Sonnet 5.5 vs Haiku 5.5: launch benchmarks, cost per feature, the opusplan pattern, and when each tier earns its price.
- [Claude Code vs Codex](https://aidenapp.org/claude-code-vs-codex) — The two leading coding agents compared by people who run both, daily, in production.
- [Claude vs GPT for coding](https://aidenapp.org/claude-vs-gpt-for-coding) — The evergreen model-family comparison: strengths, pricing, and how to decide.
## Comparisons
- [AIDEN vs Cursor](https://aidenapp.org/cursor-alternative) — Agent orchestration on a board vs an AI-first editor.
- [AIDEN vs Windsurf](https://aidenapp.org/windsurf-alternative) — Parallel worktree orchestration vs the Cascade agent IDE.
- [AIDEN vs GitHub Copilot](https://aidenapp.org/copilot-alternative) — A dedicated orchestration workspace vs Copilot inside your editor.
- [AIDEN vs Copilot coding agent](https://aidenapp.org/copilot-workspace-alternative) — GitHub's task-to-PR agent vs a local multi-agent workspace.
- [AIDEN vs Claude Code](https://aidenapp.org/claude-code-alternative) — Not a replacement, AIDEN runs your Claude Code CLI, in parallel.
- [AIDEN vs Codex CLI](https://aidenapp.org/codex-alternative) — Run OpenAI's Codex CLI inside a board-driven desktop workspace.
- [AIDEN vs Aider](https://aidenapp.org/aider-alternative) — A desktop workspace around the same idea Aider pioneered in the terminal.
- [AIDEN vs Kiro](https://aidenapp.org/kiro-alternative) — Two spec-driven IDEs: AWS's VS Code fork vs a BYOK orchestrator.
- [AIDEN vs Devin](https://aidenapp.org/devin-alternative) — A cloud AI engineer vs local agents you orchestrate yourself.
- [AIDEN vs Cline](https://aidenapp.org/cline-alternative) — The open-source VS Code agent vs a dedicated multi-agent workspace.
- [AIDEN vs Continue](https://aidenapp.org/continue-alternative) — The open-source assistant framework vs a ready-made orchestrator.
- [AIDEN vs Zed](https://aidenapp.org/zed-alternative) — Different axes: editor speed vs agent orchestration.
- [AIDEN vs v0](https://aidenapp.org/v0-alternative) — Full-codebase engineering vs Vercel's UI generation.
- [AIDEN vs bolt.new](https://aidenapp.org/bolt-alternative) — Existing codebases and real git vs in-browser app scaffolding.
- [AIDEN vs Lovable](https://aidenapp.org/lovable-alternative) — Engineer-grade workflows vs prompt-to-app prototyping.
- [AIDEN vs Replit Agent](https://aidenapp.org/replit-alternative) — Local-first orchestration vs Replit's cloud workspace.
- [Conductor alternative](https://aidenapp.org/conductor-alternative) — Conductor is a solid Mac agent runner. AIDEN adds the spec gate, the feature board, and team mode.
- [Linear alternative for AI teams](https://aidenapp.org/linear-alternative) — Linear is built for human teams. AIDEN is the board for teams where half the assignees are agents.
- [Vibe Kanban alternative](https://aidenapp.org/vibe-kanban-alternative) — Vibe Kanban shut down in April 2026. AIDEN is the maintained, spec-first board for your agents.
- [Terragon alternative](https://aidenapp.org/terragon-alternative) — Terragon shut down in February 2026. Run background agents locally instead, on your own machine.
## Category & concept guides
- [What is an agentic IDE?](https://aidenapp.org/agentic-ide) — The pillar guide: how agentic IDEs work and how they differ from chat IDEs.
- [What is an agentic development environment (ADE)?](https://aidenapp.org/agentic-development-environment) — The category above the editor: the six layers that turn raw agents into shippable work.
- [ADE vs agentic IDE vs AI IDE](https://aidenapp.org/agentic-development-environment-vs-ide) — Three terms, three scopes. What each one means and which one you actually want.
- [Local agentic development environment](https://aidenapp.org/local-agentic-development-environment) — Run agents on your own machine: BYOK, code that never leaves your filesystem, self-host.
- [Best agentic IDEs in 2026](https://aidenapp.org/best-agentic-ide) — AIDEN, Cursor, Windsurf, Devin, Kiro, Cline and Claude Code, compared honestly.
- [Desktop apps for coding agents](https://aidenapp.org/desktop-app-for-coding-agents) — The 2026 landscape: Conductor, Sculptor, the Codex app, AIDEN, and how to choose.
- [What is agentic engineering?](https://aidenapp.org/agentic-engineering) — The discipline replacing vibe coding: orchestrating agents with specs, gates, and review.
- [Vibe coding vs agentic engineering](https://aidenapp.org/vibe-coding-vs-agentic-engineering) — Same tools, opposite disciplines. Where each fits and where each breaks.
- [What is vibe coding?](https://aidenapp.org/what-is-vibe-coding) — The definition, Karpathy's origin, where it shines, and where it breaks.
- [Is vibe coding dead?](https://aidenapp.org/is-vibe-coding-dead) — What actually happened to vibe coding in 2026, and what replaced it in production teams.
- [The AIDEN Manifesto](https://aidenapp.org/manifesto) — Nine principles for building software with AI agents in 2026.
- [AI coding glossary](https://aidenapp.org/ai-glossary) — Plain-English definitions: agents, worktrees, MCP, spec gates, and more.
## Workflow & orchestration guides
- [Engineering with AI agents](https://aidenapp.org/engineering-with-ai-agents) — Why vibecoding breaks at scale, and the workflow that replaces it.
- [Claude Code orchestration](https://aidenapp.org/claude-code-orchestration) — Run several Claude Code sessions in parallel with a board, branches, and PRs.
- [Parallel agents with git worktrees](https://aidenapp.org/parallel-agents-git-worktree) — How worktrees isolate agents on separate branches, manually, then automated.
- [Multi-agent coding workflow](https://aidenapp.org/multi-agent-coding-workflow) — Design a workflow where several AI agents ship in parallel without chaos.
- [A GUI for Claude Code](https://aidenapp.org/claude-code-gui) — A desktop interface for the Claude Code CLI you already run: a board, parallel agents, a worktree per feature.
- [Manage multiple Claude Code sessions](https://aidenapp.org/manage-multiple-claude-code-sessions) — From terminal tabs to worktrees to a board: scaling past one agent without losing track.
- [Claude Code on Windows](https://aidenapp.org/claude-code-windows) — It runs on Windows today: the install, the native-versus-WSL2 call, and where AIDEN's Windows build stands.
- [Claude Code and Codex together](https://aidenapp.org/claude-code-and-codex-together) — The dual-agent workflow: which tool takes which task, and how to run both on one project.
- [Claude Code Agent Teams](https://aidenapp.org/claude-code-agent-teams) — Anthropic's experimental multi-session orchestration, and when you need more than it.
- [AI PR a