Claude Platform

Claude Platform is Anthropic's API and console for building apps and agents on Claude models, billed per token with no seat plans.

Claude Platform is Anthropic's developer product for building applications and agents on the Claude models, through one API, one console and one set of SDKs. It gives engineering teams direct access to the models, with the billing, data terms and deployment controls a production system needs. Anthropic builds and runs it, and remains an independent AI research company incorporated as a public benefit corporation. The Claude assistant at claude.ai is a separate product under the same brand, and Claude Code and Cowork belong to those subscription plans rather than to this platform. The Messages API is the center of the platform. It carries tool use, vision image input, structured outputs with constrained decoding, extended thinking, prompt caching and the Message Batches API for work that can wait. Server-side tools add web search and code execution, the Files API holds documents the model reads, and the MCP connector reaches tools over the Model Context Protocol. Managed Agents, currently in beta, runs agents on Anthropic infrastructure with permission policies, memory stores, multi-agent orchestration and self-hosted sandboxes for teams that must execute tools themselves. The Claude Agent SDK and the ant CLI define those agents in code, and the Claude Console is where organizations, workspaces and API keys are managed. The platform suits developers and product teams putting Claude inside their own software, from a first prototype to a system serving millions of requests. It fits enterprises that need zero data retention agreements, United States inference residency, or a committed contract negotiated with sales. Startups and research groups take the same models on the same rates without a contract. Usage is metered per million input and output tokens by model, with no seats and no subscription tiers, and every model, server-side tool and agent session runtime carries a published rate of its own. In daily use the platform sits behind support automation, document extraction, coding agents, research assistants and the batch jobs that run overnight. Teams reach it directly over HTTPS, through the official SDKs, or through Amazon Bedrock, Google Cloud and Microsoft Foundry where procurement already runs, and AWS and Azure marketplaces meter the same usage on an existing cloud bill. It connects outward to internal MCP servers and to any tool an application defines. Anthropic serves no embedding model of its own and points teams to Voyage AI for that half of a retrieval stack.