The IDE that codes, remembers,
and understands.
Neuron is an AI IDE built on VS Code. Use any model, keep full control of your data, with persistent codebase memory and deep code intelligence — free with your own API key.
Any LLM · Full Privacy · All VS Code Extensions
Every other AI IDE starts blind. Neuron asks the right questions from day one — so the AI understands your project before you write a single line.
Neuron walks you through a smart setup wizard. Your answers are saved to a project file the AI reads on every session.
Open any existing codebase. Neuron indexes it in the background while asking smart questions to fill the gaps.
Paste a URL and Neuron fetches the existing HTML, extracts structure and styles, and uses it as the starting point for your rebuild.
Tree-sitter powered. Neuron parses call graphs, type hierarchies, and predicts what breaks before you change it — across 20+ languages. Not keyword search. Actual code intelligence.
Neuron automatically routes simple tasks to cheap models and complex tasks to your best model — even on your own API key. It learns your patterns over time, so routing gets smarter the more you use it.
Neuron reads your project's manifest — composer.json, package.json, and more — detects the frameworks and libraries you actually depend on, pulls in their documentation, and indexes it straight into Engram memory. No stale training data, no hallucinated APIs — just the docs for the versions you're actually running.
Neuron's memory system gives the AI persistent knowledge about your project across sessions. It remembers what you decided, what you built, and why — so you never repeat yourself.
- Three-tier architecture — Sensory → Working → Long-term graph
- Auto consolidation — Deduplicates and decays stale memories
- Hybrid retrieval — BM25 full-text + semantic vector search
- Knowledge graph — Subject-predicate-object fact triples
- Memory travels with your repo — Project memories live in .neuron/engine.db alongside your code. Clone the repo and your teammate inherits the full memory context automatically — architecture decisions, patterns, everything. Global memories persist across all your projects separately.
Not grep. Not regex. Neuron parses actual ASTs with tree-sitter so the AI knows your call graph, type hierarchy, and impact radius — and Engram memory keeps what it learns: globally across all your projects, and per-project in .neuron/ inside each repo.
- Functions & classes — every symbol
- Call sites — who calls whom
- Type hierarchies — traits & interfaces
- Import graph — all dependencies
- Impact analysis — predict what breaks
- Caller tracing — across the codebase
- Callee tracing — full call chains
- Semantic search — by meaning
- Your preferences & patterns
- Cross-project conventions
- Follows you into every repo
- Architecture decisions & facts
- Saved inside the repo folder
- Clone it — teammates inherit it
Neuron's inline autocomplete predicts what you'll type next. Powered by Groq for near-instant latency — unlimited on the free tier.
Select any code, press Cmd+K, and describe your change. Neuron applies it instantly with a diff you can accept or reject.
Three modes for three workflows. Agent Mode executes autonomously. Gather Mode collects context before acting. Normal chat explains without touching files.
Every AI IDE resets when the session ends. Neuron doesn't.
| Feature | Neuron | C***** | W******* |
|---|---|---|---|
| Persistent memory across sessions | ✓ | ✗ | ~ |
| Memory travels with repo (.neuron/) | ✓ | ✗ | ✗ |
| AST-level code intelligence | ✓ | ~ | ~ |
| Smart routing on your own API key | ✓ | ✗ | ✗ |
| Auto-fetched framework docs | ✓ | ✗ | ✗ |
| Free tier — BYOK, unlimited requests | ✓ | ✗ | ✗ |
| Unlimited tab completions (free) | ✓ | ✗ | ✓ |
| Bring your own API key — free forever | ✓ | ✓ | ✗ |
| Full VS Code extension support | ✓ | ✓ | ~ |
Every AI IDE that can edit files and run commands has to make trust decisions. Here's exactly how we do it — no marketing spin.
Every tool call — file edits, shell commands, code execution — is classified into a risk tier before it runs. Higher-risk actions always require your explicit approval. Nothing runs behind your back.
Every agent action, tool call, and memory access is logged and queryable. If the AI did something, you can see exactly what and why.
Like every AI coding tool today, Neuron's approval system is a permission gate, not a hardened execution sandbox. We're not going to pretend otherwise — and we're actively working on deeper isolation. Track progress on GitHub.
Neuron doesn't route your code through a private backend like Cursor or Windsurf. Connect directly to any provider — or run models locally with Ollama.
Host DeepSeek, Llama, Qwen, and more locally. Never run out of credits.
Direct connection. Claude 4, GPT-4o, Gemini 2.5, Grok 3 — BYOK, unlimited.
Free tier beats every competitor. Paid tiers unlock frontier models — no key management required.
- ~300KTBD entry-tier tokens/month
- 10 projects indexed (hard cap)
- Private, standard usage terms
Runs on provider-hosted free endpoints. Requests may be logged by the provider for security and product improvement. Learn more →
- Unlimited usage on free, open-source models (e.g. NVIDIA's free endpoint)
- Provider-hosted — not private by default
Bring your own key. No credit card required.
- Unlimited AI requests, bring-your-own-key
- Project indexing included — embeds using your own configured model
- No credit card required
Separate from the AI usage tiers above — this covers the right to use the Neuron editor itself, annually or once.
| License | Annual | Lifetime — recommended |
|---|---|---|
Non-commercial, unlicensed Nag reminder shown on launch | $0 | — |
Non-commercial, perpetual | $20/yr | $59 |
Commercial — 1 device | $49/yr | $149 |
Commercial — 2 devices | $69/yr | $199 |
Additional seat | $20/yr | $59 |
Lifetime pricing runs roughly 3x the annual rate — it breaks even against paying annually at around year 3.
Lifetime pricing is a launch offer — available for our first 1,000 users, or through launch — exact cutoff pendingTBD. Anyone who buys in keeps their price forever; the lifetime rate may rise for new buyers once the offer ends.
What a paid license grants: Removes reminders / grants commercial useTBD — copy pending final confirmation, flagged for review.
More headroom, still on managed models.
- Unlimited project indexing
- ~4.1M entry-tier tokens and ~240K mid-tier tokens/month
- ✕No frontier model access at this tier
- Framework docs auto-fetch
- One-click deploy & test
Frontier models without managing API keys.
- Everything in Starter
- Unlimited project indexing
- Agent Mode + Gather Mode
- ~4.7M entry, ~640K mid, ~200K frontier tokens/month — mix and match, shared monthly budget
- Frontier models via Neuron, no key needed
- Checkpoints, lint auto-detection, priority speed
Our highest limits. For power users who ship fast.
- Everything in Pro
- Unlimited project indexing
- ~10.5M entry, ~1.9M mid, ~910K frontier tokens/month — mix and match, shared monthly budget
- Early access to new features
- Priority support
Token allowances are approximate. Usage draws from a shared monthly budget across entry, mid, and frontier models — mix and match freely.
The unlimited free lane runs on free, open-source models hosted by third-party providers (for example, NVIDIA's free inference endpoint) rather than on our own managed infrastructure. Because these are provider-hosted, free endpoints, the provider — not Neuron — may log requests made through this lane for security monitoring and product improvement, under that provider's own terms.
"Open source" describes the model, not the handling of your data — this lane is not private by default, and we do not recommend it for proprietary or sensitive code. For private usage, use the Limited (Managed Models) lane, a paid tier, or bring your own API key.
Powered by VS Code.
Neuron is built on VS Code. Transfer your themes, keybindings, and extensions in one click. Everything you know, plus everything AI can do.
Code at the speed of thought.
An AI IDE that thinks, remembers, and understands your codebase. Built on VS Code.
Follow us on X