Donut Labs Secures $7M for New AI-Powered Crypto Browser
New York–based Donut Labs has secured $7 million in a pre-seed financing round led by Sequoia Capital, Bitkraft, and HackVC. The round also saw participation from Matrix Partners, Sky9 Ventures, Makers Fund, SonicSVM, and prominent angel backers from the Solana ecosystem. The funding will accelerate development of Donut, the company’s innovative agentic crypto browser that uses embedded artificial intelligence agents to autonomously interact with blockchain applications.
What Is an Agentic Crypto Browser?
Unlike conventional browsers, which render HTML and execute JavaScript, Donut’s architecture integrates AI modules that understand context, user intent, and on-chain protocols. The browser features:
- Native Crypto Wallet: Secure key management with hardware isolation and multi-party computation (MPC).
- Decentralized Exchange (DEX) Integration: Direct liquidity aggregation across Uniswap, SushiSwap and layer-2 DEX protocols.
- On-chain Execution: Autonomous contract invocation powered by AI agents fine-tuned on smart contract ABIs.
- Decentralized Network Access: Peer-to-peer node discovery with gRPC and JSON-RPC fallback for high availability.
Key Technical Specifications
AI Model and Inference
Donut incorporates a transformer-based model optimized for transaction intent parsing. The model is trained on millions of smart contract interactions and on-chain event logs, enabling:
- Automated calldata translation into human-readable instructions.
- Intent classification, e.g., token swap, yield farming, NFT minting.
- Risk scoring using real-time on-chain analytics and external oracles.
Protocol Stack
The browser runs a modular stack:
- Rendering Engine: Chromium fork with WebAssembly extensions for EVM bytecode analysis.
- AI Agent Layer: On-device inference using WebNN and optional cloud offload for larger models.
- Blockchain Bridge: Secure enclave for JSON-RPC signing, threshold signatures, and gas estimation algorithms.
Security and Compliance Considerations
Donut Labs has prioritized transaction safety and regulatory compliance:
- Architecturally Isolated Signing Environments: Each transaction is signed in a secure enclave protected by Intel SGX or ARM TrustZone.
- AI-Powered Risk Screening: Continuous contract analysis against known exploit patterns and malicious addresses.
- Compliance Modules: Optional KYC/AML integration for regulated markets, built on zero-knowledge proofs to preserve privacy.
“AI translates opaque calldata into plain English, assigns safety grades, and even suggests optimized gas strategies,” said founder and CEO Chris Zhu.
Agentic Browsers Could Reshape the Internet
While initial users will likely be DeFi traders, builders, and NFT collectors, agentic browsers may enter mainstream use within five years. Zhu forecasts that leading layer-2 networks on Ethereum will leverage agentic interfaces to reduce settlement costs and that incumbent browsers like Chrome and Firefox may integrate similar AI orchestration layers or interoperate with purpose-built solutions.
In the next decade, the distinction between on-chain and off-chain applications is expected to blur entirely. Agentic browsers could become the primary entry point to the internet, akin to the shift from feature phones to smartphones.
Market Implications and Competitive Landscape
Donut enters a rapidly evolving arena:
- Existing Web3 Browsers: MetaMask, Brave, and Opera are adding wallet features but lack deep AI automation.
- AI Orchestration Startups: Competitors like Agoric and OpenAI’s hypothetical Web3 agents are exploring similar territory.
- Regulatory Environment: EU’s MiCA and forthcoming SEC guidance on digital asset custody could influence adoption pace.
Future Roadmap and Additional Use Cases
Beyond trading and yield optimization, Donut plans to expand into:
- Payments: Instant stablecoin settlements with minimal slippage.
- Gaming and NFTs: Automated minting and marketplace arbitrage.
- Content Creation: AI agents that manage royalties and distribution on decentralized platforms.
- Social Interaction: Decentralized identity (DID) integration and confidential computing for private messaging.
Technical Architecture and Protocol Stack
This new section dives deeper into Donut’s low-level design:
Node Connectivity and Data Indexing
Donut runs lightweight full nodes using Erigon for Ethereum and Paris for Solana, with on-device subgraph indexing powered by The Graph protocol. This ensures sub-second response times for balance queries and market data.
Extensibility via SDK
Developers can build custom AI agents or plugins using the Donut SDK, which exposes APIs for smart contract interaction, AI inference control, and risk-screening customization.
Security and Compliance Deep Dive
- Threshold Signatures: Multi-party computation splits private keys across devices, reducing single-point failure risks.
- On-Device Confidential Computing: Integrates AMD SEV and Intel TDX for encrypted state processing.
- Audit and Certification: Ongoing audits by Quantstamp and Trail of Bits, with SOC 2 Type II compliance in progress.
Outlook and Final Thoughts
By converging AI orchestration with blockchain execution, Donut Labs aims to redefine how users engage with digital assets and decentralized applications. As transaction costs fall and AI inference becomes more efficient, agentic browsers may well become the default gateway to Web3 and beyond.
Related: Metaplanet’s Bitcoin ‘premium’ nears $600K per BTC
Related: Polygon-backed, high-yield blockchain launches for institutional adoption