One Endpoint for
All AI Providers
One endpoint for every AI provider. Auto-fallback, cost optimization, and token compression — runs entirely on your machine.
Interactive diagram visible on desktop
How Nerve Works
From your terminal to the best AI provider in milliseconds — here's exactly what happens under the hood.
Send Request
Your CLI tool (Claude Code, Codex, Cursor) sends a request to Nerve's single endpoint. Just change the base URL — no SDK changes needed.
Nerve Hub
The request hits Nerve's local server. Auth, rate limits, and caching are checked before anything is forwarded upstream.
Task Detection
Nerve analyzes the prompt and detects the task type — coding, analysis, vision, summarization, creative, or chat — automatically.
Smart Routing
Based on the task type, Nerve picks the optimal auto/* intent, checks provider health, and routes to the best available provider with auto-fallback.
Response & Compression
The provider responds. Nerve applies 12-engine token compression (89% avg savings), tracks usage/cost, and streams the response back to your CLI.
Send Request
Your CLI tool (Claude Code, Codex, Cursor) sends a request to Nerve's single endpoint. Just change the base URL — no SDK changes needed.
Nerve Hub
The request hits Nerve's local server. Auth, rate limits, and caching are checked before anything is forwarded upstream.
Task Detection
Nerve analyzes the prompt and detects the task type — coding, analysis, vision, summarization, creative, or chat — automatically.
Smart Routing
Based on the task type, Nerve picks the optimal auto/* intent, checks provider health, and routes to the best available provider with auto-fallback.
Response & Compression
The provider responds. Nerve applies 12-engine token compression (89% avg savings), tracks usage/cost, and streams the response back to your CLI.
Request flows in
Nerve routes intelligently
Compressed response back
Powerful Features
Everything you need to manage your AI infrastructure in one place, built for scale.
Task-Aware Smart Routing
Auto-detects 7 task types (coding, analysis, vision, summarization, background, creative, chat) and routes each to the optimal auto/* intent — no manual model selection needed.
7 task types detected automatically
MCP Server Registry
24 curated MCP servers across 11 categories. Project-type detection recommends the right servers. One-click install with auto-generated config for Claude, Cursor, and VS Code.
24 servers, 11 categories, one-click install
Personalized Dashboard
Provider health leaderboard, top models, recent activity, API key usage, and daily cost charts — all with a time-range selector and real-time updates.
Real-time cost & health monitoring
Switch Models Mid-Conversation
Set a smart combo as your default, then override it per-session or per-turn — without leaving your CLI. Nerve rebuilds the cache and reverts automatically.
A live session with the hermes CLI routed through Nerve.
Connect Any Tool to Nerve
15 CLI agents, IDEs, and MCP clients have dedicated setup commands. Just point your tool at Nerve's endpoint — no SDK changes needed.
CLI Agents
7Connect via custom base URL
IDEs & Extensions
6Connect via custom base URL
MCP Clients
6Connect via MCP server endpoint
How Tools Connect
{
"baseUrl": "http://localhost:20128/v1",
"apiKey": "your-nerve-api-key"
}Built for Serious AI Workloads
Production-grade features that make Nerve more than a simple proxy — it's a complete AI gateway platform.
Smart Combos
Set a model to auto or build your own combo. 19 routing strategies with tier-1/2/3 fallback — quota out on one provider, switch to the next in milliseconds.
Memory
Persistent conversational memory — FTS5 keyword + Qdrant vector recall, built in. Your agents remember context across sessions.
3-Layer Resilience
Fail at the right level — never the whole thing. Circuit breaker per provider, cooldown per connection, lockout per model. Other keys and models keep serving.
Free Quota Pool
Pool-deduped, honest counting across 81 free-tier providers. Maximize your free tier usage without duplicate requests.
Token Compression
14 engines including RTK and Caveman, stacked on tool output and context. Save tokens and money on every request.
MCP Server
Built-in server exposing the gateway itself — 104 tools across 31 scopes, 3 transports (stdio, SSE, Streamable HTTP).
A2A Protocol
JSON-RPC agent protocol — connect an agent, it drives Nerve. 6 built-in skills including smart-routing, cost-analysis, and health-report.
Docker Prod Deployment
Containerized production deployment with bounded memory — no more dev server swap thrashing. Auto-restart timer keeps the dev server lean. 5x RAM reduction vs unmanaged dev server.
Multiple Ways to Install
Nerve runs wherever you need it — CLI, Docker, or desktop app.
Run instantly, zero install
$ npx nerveGet Started in 30 Seconds
Install Nerve, configure your providers via web dashboard, and start routing AI requests.
Install Nerve
Run npx command to start the server instantly
Open Dashboard
Configure providers and API keys via web interface
Route Requests
Point your CLI tools to http://localhost:20128
> Server running on http://localhost:20128
> Dashboard: http://localhost:20128/dashboard
> Ready to route! ✓
macOS/Linux: ~/.nerve/db.json
Windows: %APPDATA%/nerve/db.json
How Nerve Works
Data flows seamlessly from your application through our intelligent routing layer to the best provider for the job.
1. CLI & SDKs
Your requests start from your favorite tools or our unified SDK. Just change the base URL.
2. Nerve Hub
Our engine analyzes the prompt, checks provider health, and routes for lowest latency or cost.
3. AI Providers
The request is fulfilled by OpenAI, Anthropic, Gemini, or others instantly.
Powerful Features
Everything you need to manage your AI infrastructure in one place, built for scale.
Unified Endpoint
Access all providers via a single standard API URL.
Easy Setup
Get up and running in minutes with npx command.
Model Fallback
Automatically switch providers on failure or high latency.
Usage Tracking
Detailed analytics and cost monitoring across all models.
OAuth & API Keys
Securely manage credentials in one vault.
Cloud Sync
Sync your configurations across devices instantly.
CLI Support
Works with Claude Code, Codex, Cline, Cursor, and more.
Personalized Dashboard
Provider health leaderboard, top models, recent activity, API key usage, and daily cost charts — all with a time-range selector and real-time updates.
Ready to Simplify Your AI Infrastructure?
Join developers who are streamlining their AI integrations with Nerve. Open source and free to start.