Maintained reference
AI software by role and processing boundary.
This directory separates runtimes, interfaces, hosted assistants, gateways, and coding tools. It is a selection aid, not a ranking or endorsement.
Official sources checked 2026-10-24. No independent benchmark or hands-on compatibility claim is made unless a guide says so.
How to read the directory
“Local” describes where model inference is intended to run, not a guarantee that downloads, telemetry, extensions, or configured providers stay offline. Operating-system support, model compatibility, account plans, API billing, and software or model licenses are separate checks.
| Software | Role | Processing | OS | Models and licenses | Cost distinction | Official source |
|---|---|---|---|---|---|---|
| Ollama | Local runtime | Local | Windows, macOS, Linux | Packaged library; model licenses vary | Free software; verify model terms | Documentation |
| LM Studio | Desktop runtime/server | Local or configured remote | See current docs | Model support/licenses vary | Vendor terms; provider costs may apply | Documentation |
| Jan | Desktop AI app | Local or configured remote | See current docs | Model support/licenses vary | Open source; provider costs may apply | Documentation |
| GPT4All | Desktop app/local API | Local | See current docs | Supported local models; licenses vary | Open source; verify model terms | Documentation |
| llama.cpp | Inference runtime | Local | Build/backend dependent | GGUF-focused; model licenses vary | Open source | Documentation |
| LocalAI | Self-hosted API serving | Local/private infrastructure | Server support varies | Backend/model support varies | Open source; operator pays compute | Documentation |
| vLLM | High-throughput serving | Self-hosted infrastructure | Documented server environments | Architecture support varies | Open source; operator pays compute | Documentation |
| Open WebUI | Model interface | Local/cloud/hybrid by configuration | Browser plus server | Depends on backend | Open source; provider costs may apply | Documentation |
| AnythingLLM | Workspace/document interface | Local/cloud/hybrid by configuration | See current docs | Depends on backend | Free/commercial offerings vary | Documentation |
| LibreChat | Multi-provider chat interface | Cloud/hybrid by configuration | Self-hosted web app | Depends on provider | Open source; API costs may apply | Documentation |
| LiteLLM | API gateway/SDK | Cloud/local/hybrid routing | Server/application | Depends on backend | Open source/commercial options | Documentation |
| OpenRouter | Hosted API router | Cloud | API clients | Catalog changes | Usage-based; verify pricing | Documentation |
| ChatGPT | Hosted assistant | Cloud | Web/supported apps | Provider-managed | Chat plans and API billing differ | Documentation |
| Claude | Hosted assistant | Cloud | Web/supported apps | Provider-managed | Chat plans and API billing differ | Documentation |
| Gemini | Hosted assistant | Cloud | Web/supported apps | Provider-managed | Plans/API terms vary | Documentation |
| Microsoft Copilot | Hosted assistant family | Cloud | Supported Microsoft surfaces | Provider-managed | Product/account terms vary | Documentation |
| Perplexity | Search-oriented assistant | Cloud | Web/supported apps | Provider-managed | Plans/API terms vary | Documentation |
| Mistral Le Chat | Hosted assistant | Cloud | Web/supported apps | Provider-managed | Plans/API terms vary | Documentation |
| DeepSeek | Hosted assistant/model provider | Cloud; models may exist elsewhere | Availability varies | Model licenses vary | Verify service/API terms | Documentation |
| Aider | Terminal coding assistant | Configured endpoint | Terminal | Provider/local endpoint dependent | Open source; model costs may apply | Documentation |
| Continue | Editor coding assistant | Configured endpoint | Supported IDEs | Provider/local endpoint dependent | Open source/service options | Documentation |
| Cline | Editor coding agent | Configured endpoint | Supported IDE | Provider dependent | Open source; model costs may apply | Documentation |
| Roo Code | Editor coding agent | Configured endpoint | Supported IDE | Provider dependent | Open source; model costs may apply | Documentation |
| Cursor | AI code editor | Primarily cloud; verify settings | Vendor-supported desktop OSes | Managed/configured models vary | Commercial; verify plans | Documentation |
| GitHub Copilot | Coding assistant | Cloud | Supported editors/GitHub | Provider-managed | Subscription terms vary | Documentation |
| Claude Code | Terminal coding agent | Cloud with local tool access | See current docs | Provider-managed | Subscription/API terms vary | Documentation |
| ComfyUI | Node-based creative workflow UI | Usually local | Hardware/Python/node dependent | Image models/nodes have own licenses | Open source; verify every model/node | Documentation |
Selection method
- Define the task and a pass/fail check.
- Classify the data and allowed processing boundary.
- Confirm current OS, endpoint, model, license, and account or API terms in the official source.
- Run a small non-sensitive example and record version, settings, failures, and observed result.
- Recheck after any model, runtime, provider, or policy change.
Creative workflow software such as ComfyUI appears here for completeness but is referenced in a guide only when image or node-based creative work is substantively in scope.