Void: Open-source Cursor alternative
Models, Providers, and Costs
- Void supports bring-your-own-keys, including OpenRouter and Gemini; it connects directly to providers rather than proxying via its own backend.
- Commenters debate OpenRouter-style aggregators: pros (higher rate limits, one key for many models, redundancy) vs cons (5%+ markup, “just go direct” if you’re all-in on one lab).
- Users request built‑in cost tracking for BYO keys; maintainers say they already store per-model pricing and plan approximate cost displays, but tokenization differences make estimates inexact.
Forking VS Code vs Extensions/Other IDEs
- Large subthread asks why fork VS Code instead of shipping an extension like Cline/Continue.
- Extension-API limitations cited: can’t build Cursor-style inline edit boxes, custom diff UIs, full control of terminals/tabs, onboarding flows, or reliably open/close panels.
- Others argue VS Code’s restrictions are what keep its extension ecosystem fast and maintainable; a more liberal fork risks Atom/old-Firefox-style bloat and incompatibility.
- Some suggest Theia or Zed as safer long-term bases to avoid Microsoft lock‑in; others note Theia’s low adoption and Zed’s different architecture.
Void’s Feature Set and Roadmap
- Void aims to match major AI IDEs: agent mode, quick edits, inline edits with Accept/Reject UI, chat, autocomplete, checkpoints, and local/Ollama models.
- Missing today: repomap/codebase summaries and documentation indexing (@docs); maintainers currently lean on
.voidrulesplus agent/Gather and may add RAG/docs crawling or MCP integrations later. - Planned: git-branch–based agent workspaces (per-agent branches/worktrees with auto-merge via LLM) and possibly more advanced versioning schemes.
Open Source, Licensing, and Business Model
- Strong concern about “BaitWare” patterns (open source → license clampdown) referencing other projects that added branding or relicensed for enterprise.
- Void is Apache‑2; maintainers explicitly state it will remain open source and that monetization will be via enterprise/hosted offerings, not locking down the core.
- Some skepticism toward YC’s many AI IDE bets and “vibe investing”, but others argue modern OSS startups aim for win‑win splits between self-hosted OSS and paid cloud.
User Experience, Platforms, and Branding
- Linux support exists (AppImage, .deb, .tar.gz) but is easy to miss; some users hit AppImage/sandbox issues and NixOS encryption errors.
- Requests for better packaging (Homebrew, clearer download links) and more detailed README/feature comparisons, especially for non‑Cursor users.
- Mixed reactions to branding: logo seen as too close to Cursor; “open‑source Cursor” label is praised for clarity/SEO by some, but others think it signals inferiority.
- A few UX nitpicks: unexpected click sounds, tiny Linux link, need for manual folder ordering, telemetry‑off‑by‑default.
Agentic Coding vs Direct LLM Use
- Debate over whether “agentic IDEs” outperform simply using an LLM in a browser/CLI and manually steering.
- Critics say wrappers can only degrade raw model capability and that seniors mostly need autocomplete and occasional refactors.
- Supporters report big wins using agent modes for multi-file refactors, test‑edit loops, large unfamiliar codebases, and async “multitasking” while they context‑switch.
- Several contrast IDE‑based agents (Cursor/Void/Zed/VS Code Agent Mode) with CLI tools (Claude Code, Aider, Plandex), noting different preferences by experience level, workflow, and domain.
Crowded Ecosystem and Comparisons
- Commenters list a long roster of AI editors, VS Code forks, extensions, and terminal agents, calling for an eval leaderboard.
- Some favor alternative stacks (Emacs+Aidermacs, vim+plugins, JetBrains, Zed, avante.nvim) and distrust startup‑maintained VS Code forks, especially after the Windsurf acquisition.
- Others welcome Void as a rare fully open-source IDE‑level option in a field dominated by proprietary forks that proxy all traffic through vendor backends.