Run a node. Submit jobs. Self-host.

curl -fsSL https://infernetprotocol.com/install.sh | sh

infernet setup

infernet "what is 2+2?"
# or pipe input
echo "summarize: water is wet" | infernet

# 1. From PowerShell, install WSL + Ubuntu (reboot when prompted)
wsl --install -d Ubuntu

# 2. Install the Windows NVIDIA driver — gives WSL CUDA without
#    needing a separate driver inside Ubuntu.
#    https://www.nvidia.com/Download/index.aspx

# 3. Open Ubuntu, then run the same one-liner:
curl -fsSL https://infernetprotocol.com/install.sh | sh
infernet setup

infernet setup

[1/8] Node.js                ✓ v22.21.1
[2/8] Ollama                 ✗ Ollama not installed
                              Will run: curl -fsSL https://ollama.com/install.sh | sh
                              Install Ollama now (will sudo)? [Y/n] y
                              ...
                              ✓ done
[3/8] Model                  Which model should this node serve?
                                1) qwen2.5:0.5b   ≈400 MB  smoke test, runs on CPU
                                3) qwen2.5:7b     ≈4.4 GB  fits 8 GB+ GPU — recommended
                              Choice [3]:
                              → pulling qwen2.5:7b via the ollama CLI...
                              ✓ verified: ollama can serve qwen2.5:7b
[4/8] Firewall (port 46337)  ✓ rule applied via ufw
[5/8] Config                 ✓ saved to ~/.config/infernet/config.json
[6/8] Identity & control plane
                              ! no identity yet — running init
                              ...
                              ✓ identity: 5d0de683…
[7/8] Provider registration  ✓ registered
[8/8] Daemon                 ✓ daemon running

✓ setup complete

infernet setup --confirm                # auto-yes every prompt
infernet setup --model qwen2.5:7b       # preselect the model
infernet setup --no-firewall            # skip firewall step (e.g. inside containers)
infernet setup --skip-pull              # leave models alone
infernet setup --backend stub           # use canned tokens (no real GPU)

# Network (default when controlPlane.url is set)
infernet "what is 2+2?"

# Force the local engine — useful for smoke testing without a network
infernet --local "ping"

# Force network — error if no control plane configured
infernet --remote "what is the capital of France?"

# Pipe input
echo "summarize this paragraph" | infernet

# Pin a specific model (overrides config)
infernet --model qwen2.5:7b "tell me a joke"

# JSON event stream — one event per line (meta, token, done)
infernet --json "ping"

infernet model list                  # show pulled models + which is active
infernet model pull qwen2.5:7b       # pull (Ollama progress bar inline)
infernet model use qwen2.5:7b        # set as engine.model in config
infernet model show                  # current backend, host, active model
infernet model remove qwen2.5:0.5b   # delete from disk (clears active if it was)

infernet init

Control-plane URL [https://infernetprotocol.com]:        ← Enter to accept
Node role (provider|aggregator|client) [provider]:        ← Enter to accept
Human-readable node name [provider@hostname]: my-vps-1
... auto-generates Nostr keypair, detects address, asks about firewall ...
Wrote /home/ubuntu/.config/infernet/config.json
Pubkey: 5d0de683a5f22aa1d5a8927a431d86601277aad61fc7cdce126ac8c012e2c84d

infernet init --url https://infernetprotocol.com \
              --role provider \
              --name my-vps-1 \
              --p2p-port 46337 \
              --no-advertise          # outbound-only mode (Tor-friendly)

infernet start                    # detach to background, return shell prompt
infernet start --foreground       # for systemd (Type=simple) or PM2

# Inspect state
infernet status                   # remote row + local stats
infernet stats                    # live in-memory snapshot via IPC
infernet logs                     # tail ~/.config/infernet/daemon.log
infernet stop                     # graceful shutdown via IPC, signal fallback

infernet doctor

[1/6] Local config           ✓ provider/my-vps-1, key=5d0de683a5f2…
[2/6] Engine                 ✓ Ollama up @ http://localhost:11434, model qwen2.5:7b pulled
[3/6] Control plane          ✓ https://infernetprotocol.com (HTTP 200, 124ms)
[4/6] Daemon                 ✓ pid=12345, last heartbeat 4s ago, jobs 12/13
[5/6] Provider row           ✓ registered as 5d0de683…, status=available, gpu=A100
[6/6] End-to-end             ✓ routed via=p2p, provider=self, first token 740ms, total 1280ms

All checks passed.

infernet doctor --skip-e2e

# Bring your own wallet (recommended — non-custodial)
infernet payout set BTC mainnet bc1q9h6zcv4fp9kx5n0jq5qf3w2yxz4z3y2w5
infernet payout set ETH mainnet 0x742d35Cc6634C0532925a3b844Bc9e7595f8B59f
infernet payout set SOL mainnet 5RaBwVPRBfpz9MRBnGRcQAPfvKxcUVtxL2ANe9XJ
infernet payout set USDC solana 5RaBwVPRBfpz9MRBnGRcQAPfvKxcUVtxL2ANe9XJ
infernet payout set USDC polygon 0x742d35Cc6634C0532925a3b844Bc9e7595f8B59f

# Lightning — paste your bolt12 (recommended) or LNURL-pay address
infernet payout set BTC ln lno1qsgqmqvgm96frzdg8m0gc6nz...

# Don't have an LN node? Opt-in to a CoinPay-custodial channel:
infernet payout set BTC ln --provision

# Inspect
infernet payout list
infernet payout remove ETH mainnet
infernet payments                    # recent transactions (signed read)

infernet tui                # default 2s poll
infernet tui --refresh 1000 # 1s — for live debugging

# Install the unit (writes ~/.config/systemd/user/infernet.service)
infernet service install
infernet service enable          # systemctl --user enable --now infernet

# Inspect
infernet service status
infernet service logs            # journalctl --user -u infernet -f

# (Optional) make it survive logout / reboot when not logged in:
loginctl enable-linger $USER

# Tear down
infernet service disable
infernet service uninstall

infernet service unit

# Send a magic link
curl -sS -X POST https://infernetprotocol.com/api/auth/login \
  -H "Content-Type: application/json" \
  -d '{"email":"you@example.com"}'

# Or password
curl -sS -X POST https://infernetprotocol.com/api/auth/login \
  -H "Content-Type: application/json" \
  -d '{"email":"you@example.com","password":"hunter22"}' \
  -c cookies.txt

# Use the resulting cookie for /api/admin/* (when those land — IPIP-0003 phase 4)

infernet gpu                           # human summary
infernet gpu --json | jq .interconnects # raw detected shape

{
  "specs": {
    "interconnects": {
      "nvlink":     { "available": true,  "topology": "all-to-all", "link_count": 6 },
      "xgmi":       { "available": false, "topology": "none",       "link_count": 0 },
      "infiniband": { "available": true,  "active_port_count": 2 },
      "efa":        { "available": false, "adapter_count": 0 },
      "rdma_capable": true
    }
  }
}

curl -sS -X POST https://infernetprotocol.com/api/chat \
  -H "Content-Type: application/json" \
  -d '{
    "messages":[{"role":"user","content":"hello"}],
    "modelName":"qwen2.5:7b",
    "maxTokens":256
  }'

# → { "jobId":"...", "streamUrl":"/api/chat/stream/..." }
# Then SSE-tail streamUrl for tokens.

curl -sS https://infernetprotocol.com/api/peers?limit=10 | jq .

# → {
#   "data":[
#     {
#       "pubkey":"5d0de683…",
#       "multiaddr":"/ip4/162.250.189.114/tcp/46337",
#       "last_seen":"2026-04-26T12:00:00Z",
#       "served_models":["qwen2.5:7b"],
#       "gpu_model":"A100"
#     }
#   ]
# }

curl -sS https://infernetprotocol.com/.well-known/did.json | jq .verificationMethod

# Use the publicKeyMultibase to verify any Receipt signed by
# did:web:infernetprotocol.com.

git clone https://github.com/profullstack/infernet-protocol.git
cd infernet-protocol
cp sample.env .env
node tooling/generate-secrets.mjs >> .env       # CLI session, cron, DID keys
$EDITOR .env                                    # fill SUPABASE_*, RESEND, CoinPay, etc.

pnpm install
pnpm --filter @infernetprotocol/web build

# --- Supabase ---------------------------------------------------------------
NEXT_PUBLIC_SUPABASE_URL=https://<project-ref>.supabase.co
NEXT_PUBLIC_SUPABASE_ANON_KEY=eyJ...
SUPABASE_SERVICE_ROLE_KEY=eyJ...                  # server-only, never browser

# --- Public-facing URL (used by every redirect, OG card, install.sh) -------
NEXT_PUBLIC_APP_URL=https://your-infernet.example

# --- Auth (CLI bearer JWTs, cron auth, DID:web platform identity) ----------
INFERNET_CLI_SESSION_SECRET=...                   # generate-secrets.mjs fills these
INFERNET_CRON_SECRET=...
INFERNET_PLATFORM_DID_PRIVATE_KEY=...

# --- Email (sign-up confirmation, password reset, /contact) ----------------
RESEND_API_KEY=re_...

# --- Optional: NIM fallback for the public /chat playground ----------------
NVIDIA_NIM_API_KEY=nvapi-...

# --- Payments (CoinPayPortal — non-custodial except Lightning) -------------
COINPAY_ISSUER_API_KEY=cprt_...

# --- Web server -------------------------------------------------------------
PORT=8080

# Multi-stage build — preserves the pnpm workspace topology.
docker build -t infernet-control-plane -f docker/Dockerfile .

# Run with your .env mounted in
docker run --env-file .env -p 8080:8080 infernet-control-plane

pnpm --filter @infernetprotocol/web start       # http://localhost:8080
# Front it with nginx/Caddy + Let's Encrypt for TLS, then bind a systemd
# unit so it restarts on crash. apps/cli/commands/service.js generates a
# matching unit for the GPU node side.

curl -fsS https://your-infernet.example/api/health
# {"ok":true,"uptime_s":...,"node_env":"production","commit":"<sha>"}

# Round-trip the device-code login from a workstation:
infernet init --url https://your-infernet.example
infernet login                                  # opens browser → polls
infernet "ping"                                 # default verb is chat

infernet init --url https://your-infernet.example
infernet setup                                  # installs Ollama, pulls a model, opens firewall
infernet register                               # advertises the node on your control plane
infernet start                                  # daemon takes paying jobs

┌────────────────────────────────────────────────────────┐
│  Control plane (Next.js + Supabase)                    │
│   /api/v1/node/*      provider tier  (Nostr-signed)    │
│   /api/admin/*        dashboard tier (cookie session)  │
│   /api/chat, /api/peers, /.well-known/did.json         │
│                                                        │
│   Writes CPR Receipts to coinpayportal.com on every    │
│   completed job — operator reputation accumulates      │
│   automatically.                                       │
└─────────────────────────────────────┬──────────────────┘
                                      │  (server-side only)
                                      ▼
                              Supabase / Postgres
                                      ▲
        Nostr-signed HTTP             │           anon SSE / cookie
        (BIP-340 Schnorr)             │
   ┌────────────┐   ┌──────────┐   ┌────────────────────┐
   │ infernet   │…  │ Browsers │   │ NVIDIA NIM         │
   │ daemon     │   │ /chat    │   │ (fallback only)    │
   │ Ollama     │   │ /status  │   │                    │
   │ TCP 46337  │   │ /docs    │   └────────────────────┘
   └────────────┘   └──────────┘