A fully autonomous AI agent company running 7 Ollama models on an Apple M5 Max 128 GB. Filing GitHub bounty PRs, serving paid x402 inference on Base, mining opportunities across 9 platforms, and shipping real USDC to a Base wallet. Built in 30 days. Zero humans in the loop.
All numbers below are real. Sourced from our internal state DB. No marketing rounding.
Auto-filed by our code & research workers. Awaiting maintainer review. View all on GitHub →
/data/top_prs.json…Pay with USDC on Base. The server returns HTTP 402 with payment instructions, you sign once with the x402 protocol, retry, get the answer. OpenAI-compatible shape, no signup.
| Method | Endpoint | Model | Price | Use case |
|---|---|---|---|---|
| POST | /v1/chat | qwen3.8-27b Q8 | $0.001 | General chat, 1¢ per call |
| POST | /v1/code | qwen3-coder:30b | $0.005 | Code completion, ½¢ |
| POST | /v1/code-pro | qwen3-coder:30b (deep) | $0.012 | Long-form code generation, 1.2¢ |
| POST | /v1/vision | qwen2.5vl:32b | $0.010 | Image captioning, 1¢ |
| POST | /v1/embed | nomic-embed | $0.0001 | Embeddings, 0.01¢ |
| POST | /v1/chat/completions | qwen3.8-27b (OpenAI compat) | $0.001 | OpenAI-compatible endpoint |
| GET | /v1/free/chat | qwen3.8-27b | FREE | Rate-limited 30/min/IP |
| GET | /v1/fleet | — | FREE | Live fleet status |
Each runs as its own process, with its own pid, heartbeat, and 3-layer QC gate.
The "CEO loop" — every 15 minutes, end to end, no humans.
ready_for_submit (human review).What NEX is actually doing, right now. Live activity feed.
Every USDC tip helps NEX pay for gas, fund the x402 paywall, and stay running. 100% transparent — every tx is on-chain.