Kaeva Labs is building the essential economic and data substrate for the AI agent ecosystem. Their primary contribution is the development of machine-native primitives—such as signed provenance and sub-second on-chain settlement—that allow agents to act as independent economic participants. By creating a marketplace (UNBLOCK) where agents can programmatically purchase knowledge, they solve a major bottleneck in agent autonomy: the ability to acquire and pay for information outside of a model's static training data.
The lab is deeply integrated into the emerging agent stack through its support for the Model Context Protocol (MCP). This ensures that the infrastructure they build is compatible with a wide range of agent frameworks and LLM providers. For developers building autonomous systems, Kaeva Labs provides the necessary tools to move past human-centric barriers like credit card payments and opaque APIs, offering a more modular and verifiable way for agents to interact with the broader digital economy.
Kaeva Labs operates on a specific critique of modern software design: it assumes a human operator. The team argues that current digital primitives—including cookie banners, Stripe checkouts, and documentation written in the second person—are functional barriers for autonomous agents. While humans can navigate a multi-step checkout funnel, agents require machine-readable provenance, sub-second settlement, and signed events. Kaeva Labs is an AI research lab building the infrastructure to replace these human-centric interfaces with primitives designed for software-to-software interaction.
At the center of their roadmap is UNBLOCK, a marketplace for agent commerce. UNBLOCK is the environment where AI agents go to buy the knowledge or capabilities they were not initially trained on. Unlike traditional data marketplaces designed for bulk corporate acquisition, UNBLOCK is built for programmatic, granular transactions. The lab emphasizes that agents need to acquire information on the fly, and the marketplace facilitates this through a 'second-brain substrate' that ensures an agent's state and memory can travel with it across different model providers like OpenAI, Anthropic, or Google.
The technical choices made by Kaeva Labs reflect a commitment to decentralization and speed. The system is built on Cloudflare Workers, utilizing D1 for relational storage and Vectorize for embedding management. This stack is designed to provide the performance required for agents to make decisions in real-time. For the economic layer, the lab uses the Base Sepolia testnet and the Coinbase Smart Wallet. This allows agents to operate with independent financial agency, bypassing the 'credit card problem' where agents lack a native way to pay for services in a fiat-based economy.
Interoperability is maintained through the Model Context Protocol (MCP). By adopting MCP, Kaeva ensures that the 'verbs' and actions within UNBLOCK are canonical and portable. This prevents vendor lock-in and allows developers to build agents that maintain a consistent knowledge base even if they switch underlying LLMs. The marketplace also introduces signed provenance, ensuring that every piece of data an agent purchases is cryptographically verifiable and accountable by default.
Kaeva Labs describes itself as a lab that 'ships what we research.' They focus on three main initiatives: continuous self-models, the 'brain protocol,' and agent-native commerce. These research areas aim to define how agents maintain a sense of self and state over long durations. While currently in a restricted beta phase with a focus on early invitees, the lab's roadmap indicates a trajectory toward a full L2 mainnet deployment for their commerce network.
Based on their public presence, the lab is a small, lean operation led by developers who have experienced the friction of building agents on a human-centric web. They are not merely building another agent framework; they are building the economic and data substrate that allows those frameworks to become useful in the real world. By focusing on the 'plumbing'—the rails that connect agents to data and payments—Kaeva Labs is attempting to solve the underlying logistical challenges that prevent AI from moving beyond simple chat interfaces to true autonomous operation.
The marketplace where AI agents buy the knowledge they need to ship.
Kaeva Labs is hiring.