Anthropic is the company behind the Claude family of models and the tools built on them, including the agentic coding solution Claude Code and the Claude Developer Platform. It publishes engineering guidance for teams building agents on Claude through its Engineering at Anthropic blog.
Its Applied AI team writes on building agents in practice. That team’s stated position is that Context engineering is the natural progression of Prompt engineering, and that context should be treated as a finite resource to be curated rather than filled — a view set out in Effective context engineering for AI agents. Anthropic gives “do the simplest thing that works” as its standing advice for teams building agents on top of Claude.
A post on long-running agents describes two further pieces of Anthropic’s published engineering work: a two-agent harness for autonomous full-stack development (an initializer agent that sets up a project once, and a coding agent woken up repeatedly to make incremental progress, run tests, and commit), and the Brain/Hands/Session Split architecture behind Claude Managed Agents. The same post describes Anthropic’s Project Vend, in which a Claude instance ran an actual office vending business for a month, as an early public demonstration of what happens when an agent has to maintain a coherent identity across weeks rather than single sessions, and cites a separate scientific-computing case study in which Claude Opus 4.6 built a Boltzmann solver over a few days that reached sub-percent agreement with a reference implementation.
Activities & Products
- The Claude family of models.
- Claude Code, its agentic coding solution.
- The Claude Developer Platform, on which Anthropic has released a memory tool in public beta for storing and consulting information outside the context window through a file-based system, and a tool result clearing feature.
- Claude Managed Agents, its hosted runtime for long-running agents.
- Engineering at Anthropic, the blog on which it publishes guidance for developers.