Cluster
Dev AI
Coding agents, local STT, RAG UIs, and diffusion frontends. Proprietary vs open called out plainly. Continue’s Cursor acquisition is noted on its page — no nostalgia cosplay.
- Aider
Apache-2.0 terminal pair-programmer for git repos; you bring the model.
- Continue
Apache-2.0 coding agent; acquired by Cursor — verify maintenance before you depend on it.
- Tabby
Self-hosted coding assistant. Apache-2.0 core; ee/ is separate enterprise license.
- OpenHands
MIT Agent Canvas / coding-agent platform; sandboxes and Cloud/enterprise are separate stories.
- Goose
Apache-2.0 extensible agent (install/edit/test with any LLM); MCP-capable.
- Cursor
Proprietary AI code editor (Anysphere). Subscription — not open source.
- Zed
GPL-3.0-or-later editor (+ Apache-2.0 components); hosted AI under Zed ToS/pricing.
- Cline
Open coding agent in the editor; confirm current license and model routing.
- Kilo Code
MIT coding agent for VS Code, JetBrains, and CLI; model bills still apply.
- Plandex
MIT terminal agent for large multi-file tasks; Cloud winding down — self-host + BYO keys.
- GitHub Copilot
Hosted coding assistant. Proprietary; org policy and telemetry matter.
- whisper.cpp
Local Whisper STT in C/C++. MIT on the runner; model terms still apply.
- AnythingLLM
Local-first RAG / agent UI. MIT on the open repo.
- Dify
LLM app / RAG / agent platform. Apache-2.0 with multi-tenant and logo conditions.
- Langfuse
MIT-core LLM observability (traces/prompts/evals); ee/ is Enterprise License.
- ComfyUI
Node-based diffusion graphs. GPL-3.0. VRAM-hungry.
- Stable Diffusion WebUI
AUTOMATIC1111 Gradio UI. AGPL-3.0. Check upstream cadence.
Hardware that usually matters
Local STT, RAG, and diffusion UIs are cheap to download. The box is not. VRAM first, then RAM, then disk. Example Amazon listings — not rankings. Same spine as the Local LLMs hardware notes.
- ASUS Prime GeForce RTX 5060 Ti 16GB — mid-range 16 GB VRAM.
- GIGABYTE GeForce RTX 3090 24GB — 24 GB VRAM for heavier local models and image pipelines.
- CORSAIR Vengeance DDR5 64GB (2×32GB) — system RAM for offload.
- Samsung 990 PRO 2TB NVMe SSD — model files eat disk.
Some links are Amazon Associates. We may earn a commission if you buy.