# llama.cpp b12344 cancels tool calls after disconnects

> The llama.cpp release adds server-side cancellation for MCP tool calls when a client disconnects or times out, and includes builds for several platforms.

Oossa · 2026-11-13 · https://oossa.com/en/llama-cpp-b12344-cancels-tool-calls-after-disconnects

The ggml-org team released llama.cpp b12344 with a server change that cancels MCP tool calls when a client disconnects or times out. MCP is a way for AI apps to connect to external tools. The release page also offers builds for macOS, iOS and several Linux systems.

## The facts

- The change is listed as pull request #32025.
- The release includes macOS Apple Silicon and Intel builds, an iOS XCFramework, and Linux CPU builds.

## Why it matters

For developers running llama.cpp servers, the change is intended to stop tool calls from continuing after a client has disconnected or timed out; the release notes do not say how much server capacity this may save.

## Sources & references

1. [ggml-org/llama.cpp b12344](https://github.com/ggml-org/llama.cpp/releases/tag/b12344) – llama.cpp, 2026-11-13

Last updated: 2026-11-13