Claude news November 2025
3 updates from Claude and Anthropic in November 2025, from product launches to research papers, each with a short satori. note on what it means for a European business.
Effective harnesses for long-running agents
A two-part system with an initialiser and an incremental coder. Claude keeps track across several context windows with feature lists and git.
This is the secret behind long autonomous sessions that do not lose the thread. Good reading if you plan to build something agentic for real.
Introducing advanced tool use on the Claude Developer Platform
Three beta features, dynamic tool discovery, code-based execution instead of sequential API calls and learning from examples.
Suddenly you no longer need to define 50 tool schemas, the model finds them itself. For agents it is the difference between a postcard and GPS.
Code execution with MCP: Building more efficient agents
Agents run code against MCP servers instead of making direct tool calls, and token consumption drops by up to 98.7 percent.
Almost ninety-nine percent fewer tokens. That is not optimisation, it is a whole new cost picture. For companies on a budget it can decide whether the AI project happens at all.
We use analytics cookies to see which pages people read, and marketing cookies stay off unless you tick them under Details.