From the filing
PDF page 212
Printed filing page 201 · Extracted page text
Read the page text
Table of Contents • Model Hardware Standard. Currently in research preview, our Model Hardware Standard enables AI agents to find, understand, and safely operate lab and manufacturing equipment such as microscopes, liquid handlers, and off-the-shelf robotic arms. Our Model Hardware Standard is model-agnostic and lets any agent harness reach devices with a programmable interface through standard protocols such as MCP. Infrastructure The Claude Platform offers high reliability and availability for enterprise-grade agentic workloads. Customers building on Claude need the model served dependably wherever they operate, they need it to run efficiently at scale and speed, and they need the tooling that allows a model to work on its own. We operate that infrastructure so our customers do not have to build and maintain it themselves. • • Availability. The Claude Platform is available directly from Anthropic and through the major cloud providers, including AWS, GCP, and Azure. Feature availability varies by deployment. Cost, Speed, and Scale. Developers can tune how Claude runs to fit the performance and economics of each workload: Prompt Caching. Large, reused prompt components can be cached for up to one hour, reducing cost by up to approximately 97% and latency by approximately 60% for long prompts, compared with processing the same requests without caching. Batch Processing. Enables high-volume asynchronous processing at reduced cost. Fast Mode. Delivers lower-latency responses for latency-sensitive and interactive applications where speed is critical. ◦ ◦ ◦ Harnesses A model alone is not an agent. Running one reliably in production is challenging, because agents must survive crashes and interruptions, act on systems with scoped and auditable credentials, and scale across variable workloads. Autonomous operation requires a harness, which provides the secure environment, credentials, memory, and orchestration loop a model needs to plan, call tools, verify its work, and complete a task safely on its own. Building and operating that infrastructure is specialized, continuous work that, until recently, each customer had to take on themselves. We offer two ways to run a harness on the Claude Platform: • • Agent SDK. This solution makes the same production-grade harness that powers Claude Code available to all developers, to run on their own compute. It provides built-in context management, session handling, multi-agent coordination, and prompt optimization, so customers can build and run sophisticated agents on their own infrastructure in a fraction of the time otherwise required. Claude Managed Agents. An out-of-the-box solution in which we operate the harness on our customers’ behalf, on Anthropic-managed compute. We host the sandboxes, store the credentials, maintain the memory, and run the agent loop; customers define their agent’s tasks, tools, and guardrails. When a new model becomes available, customers do not need to rebuild their agent loop. Direct Applications and Integrations Our platform primitives and models are the foundation for all of our products and support a wide range of use cases, from coding agents to specialized products in life sciences, finance, security, and more. Paired with our deep understanding of current and future capabilities of our models, this foundation allows us to rapidly deliver products that unlock value across many user types and workflows. • Claude Code. A coding agent, available as a terminal tool, desktop app, web interface, and mobile app, with native integrations for GitHub and popular IDEs (Integrated Development Environments). Claude Code handles multi-step software development tasks—planning, writing, editing, testing, debugging, and committing code—across large codebases with sustained context and supports parallel agent sessions, so a single developer can keep multiple tasks moving in parallel. 201
Text is provided for reference. Check the original for tables, images, and exact formatting.