Bring Your Own LLM
Neural Inverse connects directly to your LLM provider of choice. No proxy server. No token routing. No data leaving your machine unless you send it to the provider yourself.
20 Providers. One IDE.
Cloud API Providers
| Provider | Strength | Auth |
|---|---|---|
| Anthropic | Best coding models (Claude Opus/Sonnet) | API key |
| OpenAI | GPT-5.4, o3, o4-mini reasoning | API key |
| Google Gemini | 1M context, multimodal | API key |
| xAI | Grok-4, fast reasoning | API key |
| DeepSeek | Low-cost reasoning (DeepSeek-R1) | API key |
| Mistral | Devstral coding, Codestral FIM | API key |
| Groq | Ultra-fast open model inference | API key |
| OpenRouter | 300+ models, single API key | API key |
| GitHub Models | 40+ models from one GitHub PAT | GitHub PAT |
| Fireworks AI | Fastest open-model inference, native tools | API key |
| Cerebras | 2000+ tok/s wafer-scale hardware | API key |
Infrastructure Providers
| Provider | Use Case | Auth |
|---|---|---|
| AWS Bedrock | Enterprise, VPC-internal models | AWS credentials |
| Google Vertex AI | GCP-hosted Gemini | Service account |
| Microsoft Azure | Azure-hosted OpenAI models | Resource key |
Local / Self-Hosted
| Provider | Use Case |
|---|---|
| Ollama | Run models on your laptop (auto-detected) |
| vLLM | Production-grade local serving |
| LM Studio | Desktop model runner (auto-detected) |
| LiteLLM | Universal proxy for any backend |
| OpenAI-Compatible | Any OpenAI-format endpoint |
Per-Feature Model Selection
Different tasks need different models. Neural Inverse lets you assign independently:
| Feature | Best for | Recommended model type |
|---|---|---|
| Chat (Ctrl+L) | Complex code generation, architecture | Powerful (Claude Opus, GPT-5.4, Gemini Pro) |
| Inline Edit (Ctrl+K) | Quick targeted changes | Mid-tier (Sonnet, GPT-4.1-mini) |
| Autocomplete (Tab) | Real-time suggestions | Fast (local Ollama, Cerebras, Groq) |
| Power Mode (Cmd+Alt+P) | Autonomous multi-step agents | Powerful with tool support |
| Apply | Executing code changes | Fast (local preferred) |
Run a $0 local model for autocomplete. Use Claude Opus for Power Mode. No compromise.
Your Keys, Your Machine
- API keys stored in your OS keychain
- Network requests go directly from your machine to the provider
- No Neural Inverse servers in the path
- No telemetry on your prompts or responses
- Works fully air-gapped with local providers
Get Started
# macOS / Linux
curl -fsSL https://neuralinverse.com/sh | bash
# Windows
irm https://neuralinverse.com/win | iexOpen Settings > Neural Inverse > LLM Providers. Enter your key. Start coding.
Copyright 2026 Neural Inverse Inc.
Was this page helpful?