A verifiable intelligence graph built on OpenGradient. Every question you ask becomes a cryptographically proven reasoning chain recorded permanently on the OpenGradient blockchain.
Live: proofgraph.vercel.app Code: github.com/Cryptbits/proofgraph
Most AI tools give you an answer and ask you to trust it. ProofGraph proves the answer instead.
When you submit a question, three parallel reasoning nodes fire simultaneously inside OpenGradient's TEE. Each node approaches the problem from a distinct angle. Core Analysis breaks down the technical mechanisms. Evidence and Context brings the real-world facts and comparisons. Key Takeaways gives the practical implications. All three run at the same time, each inside a hardware-sealed AWS Nitro Enclave, each paying for itself autonomously via x402 in $OPG on Base Sepolia, each producing its own transaction hash on the OpenGradient Nova Testnet.
Once all three complete, a fourth node — the Final Synthesis — takes all three outputs and combines them into one clear, structured, verified answer. That is four TEE inference calls, four transaction hashes, and four nodes on the graph per query.
Every node appears live on a D3.js force graph as it completes. Every node is clickable. Every node links to explorer.opengradient.ai so the proof can be independently verified by anyone.
The graph grows over time. New questions find and cite verified nodes from previous sessions, building a permanent tamper-proof record where knowledge compounds across queries instead of resetting with every session.
TEE Inference
All four inference calls run via og.LLM inside AWS Nitro Enclaves. The hardware proves each computation ran correctly and the output hash settles on-chain. No trust in any company required.
x402 Payments Each of the four inference calls is paid per-request in $OPG via Permit2 on Base Sepolia, embedded directly inside the TEE. No subscriptions, no middleware, no human approval needed between the request and the compute.
Digital Twin Routing Each of the three reasoning nodes routes to the most relevant specialist persona from Twin.fun based on the question topic. Technical questions go to an AI Research Expert. DeFi and protocol questions go to a DeFi Analyst. Investment and economics questions go to Naval Ravikant. Each twin's perspective shapes the node output before the synthesis step.
MemSync Completed sessions are stored in OpenGradient's portable memory layer, giving the graph persistent cross-session context that carries forward with every new query.
Nova Testnet Every proof hash from every node settles on the OpenGradient Nova Testnet. Every node detail panel in the UI links directly to the block explorer for public verification.
You need Python 3.11+, $OPG testnet tokens from faucet.opengradient.ai, and Base Sepolia ETH for Permit2 gas from alchemy.com/faucets/base-sepolia.
Clone the repo and run setup:
git clone https://github.com/Cryptbits/proofgraph
cd proofgraph
bash setup.shAdd your wallet private key to backend/.env:
OG_PRIVATE_KEY=0x_your_private_key_here
Start the backend in one terminal:
source venv/bin/activate
cd backend && python main.pyStart the frontend in a second terminal:
cd frontend && python3 -m http.server 3000Open http://localhost:3000.
Backend on Render
Connect the repo to render.com, set OG_PRIVATE_KEY in the environment variables, and Render handles the rest via the Procfile.
Frontend on Vercel
Connect the repo to vercel.com, set BACKEND_URL in frontend/index.html to your Render service URL, and push. Both platforms redeploy automatically on every push to main.
proofgraph/
backend/
main.py FastAPI server, WebSocket streaming, event buffering
graph_engine.py 3-node parallel pipeline and synthesis
og_client.py OpenGradient SDK wrapper using og.LLM and Permit2
og_knowledge.py OG knowledge base for offline fallback
twin_router.py Digital Twin routing engine
memsync_client.py MemSync memory storage
database.py SQLite persistence
models.py Pydantic data models
frontend/
index.html Single-file UI with D3.js graph and WebSocket streaming
| Layer | Technology |
|---|---|
| Verifiable inference | OpenGradient TEE via og.LLM |
| Payments | x402 with $OPG on Base Sepolia |
| Memory | OpenGradient MemSync |
| Digital Twins | Twin.fun persona routing |
| Proof settlement | OpenGradient Nova Testnet |
| Frontend | D3.js, vanilla JS, WebSockets |
| Backend | FastAPI, Python, SQLite |
Never commit your private key. The backend/.env file is in .gitignore. For production deployments, set OG_PRIVATE_KEY exclusively through your hosting platform's environment variable settings and never in code or config files.
opengradient.ai · explorer.opengradient.ai · faucet.opengradient.ai · docs.opengradient.ai · x.com/OpenGradient · discord.gg/2t5sx5BCpB