Fabric by Carmel Labs

Another key feature group is the platform's economic model, which includes per-second billing with no cloud markup, dramatically reducing costs for users running AI inference jobs. This pricing structure enables developers to access compute power at a fraction of the cost of traditional cloud services, making AI experimentation and deployment more financially feasible. Simultaneously, device owners can earn passive income by contributing their idle compute resources to the network, creating a symbiotic ecosystem where both supply and demand sides benefit. The cryptographically verified workloads ensure that all computations are secure and tamper-proof, maintaining integrity and trust within the distributed network, which is essential for handling sensitive AI tasks and data.

Fabric also provides robust technical capabilities, such as seamless integration with popular AI frameworks and tools, including OpenAI, Anthropic, LangChain, CrewAI, and MCP, facilitating easy adoption for developers already using these ecosystems. The platform supports continuous monitoring and validation through its sister product, AgentStatus, which leverages Fabric's infrastructure to probe AI agents from real devices and networks, offering features like LLM-as-Judge semantic validation and gold prompt testing with report cards. This integration ensures that AI agents are not only cost-effective to run but also reliable and correctly functioning, with alerts via Slack, PagerDuty, or webhooks to notify users of any issues, thereby enhancing operational resilience and user confidence in agent performance.

Technically, Fabric operates by aggregating idle compute power from a decentralized network of verified consumer devices worldwide, using cryptographic methods to authenticate devices and secure workloads. The platform manages job distribution, scheduling, and execution across these nodes, ensuring efficient utilization of resources while maintaining low latency and high availability for AI inference tasks. It employs residential IPs to provide authentic network conditions, which is critical for realistic AI agent testing and deployment, distinguishing it from traditional cloud providers that rely on data center IPs. This approach enables scalable, distributed computing that mirrors real-user environments, offering a practical solution for AI workloads that require geographic diversity and cost efficiency.

Users benefit from measurable outcomes such as significant cost savings on AI inference, often at a fraction of traditional cloud expenses, due to per-second billing and the elimination of cloud markup. They gain access to a globally distributed network with residential IPs, improving the realism and reliability of AI agent testing, which leads to higher agent performance and user satisfaction in production. The platform also provides passive income opportunities for device owners, creating an additional revenue stream from unused computational resources. Furthermore, the integration with monitoring tools like AgentStatus ensures continuous validation and quick issue resolution, reducing downtime and enhancing the overall trustworthiness of AI applications in real-world scenarios.

Concrete use cases include AI developers running large-scale inference jobs for language models or image generation, where Fabric's cost-effective compute power allows for extensive experimentation without prohibitive expenses. Another example is companies deploying AI agents for customer service or automation, using the residential IP network to test agent responses from various global locations, ensuring they work correctly for diverse user bases. Researchers can leverage the platform for distributed AI training or simulation tasks, benefiting from the scalable and geographically diverse infrastructure. Additionally, device owners, such as individuals with idle PCs, can contribute to the network to earn income, while organizations use Fabric alongside AgentStatus for continuous monitoring and validation of their live AI agents.

The target users include AI developers, data scientists, and engineers building and deploying AI agents who need affordable, scalable compute resources with realistic network conditions. It also appeals to companies and startups implementing AI solutions for automation, customer interaction, or analytics, seeking to reduce infrastructure costs and improve agent reliability. Integrations are supported with major AI frameworks like OpenAI, Anthropic, LangChain, CrewAI, and MCP, as well as alerting systems such as Slack and PagerDuty. The tech stack involves distributed computing, cryptographic verification, and residential IP networking. Pricing plans are based on per-second billing without cloud markup, making it accessible for various budget levels, from individual developers to large enterprises.

In summary, Fabric by Carmel Labs provides a transformative distributed compute platform that democratizes access to AI infrastructure by connecting idle device power with AI workloads at reduced costs. It addresses key challenges in AI deployment, such as high expenses and lack of realistic testing environments, through a global network of verified residential devices. By enabling both cost savings for users and income for contributors, Fabric fosters a sustainable ecosystem that enhances the reliability and accessibility of AI technologies, solidifying its role as a critical trust layer in the evolving AI agent economy.

Updated 2026-02-28