The product includes a MCP server with 15 tools accessible through the Model Context Protocol, allowing AI agents to call them directly, backed by full documentation and benchmarks at the provided URLs.

Benchmarks report a 31-for-31 success rate across seven languages, with 83 real-world wins and zero regressions, drawn from nearly 5,000 file merges and over 1,500 downloads across languages like C, Python, and Go.

Weave relies on tree-sitter to parse code and performs merges by function and class, supporting 28 programming languages plus five data formats, including languages like TypeScript, JavaScript, Python, Go, Rust, Java, C, and more.

Weave is an entity-level semantic merge driver for Git that resolves conflicts by merging at function and class levels rather than line-by-line, enabling cleaner integrations.

When two agents edit separate functions within the same file, weave avoids traditional line-level conflicts and results in a clean merge, unlike standard Git merges that often show conflicts in files such as src/lib.ts.

Weave offers three integration layers: using the merge driver on its own, coordinating multi-agent workflows, and a CRDT-based state where agents claim entities before editing to detect conflicts proactively.

Practical usage examples cover installing weave via Homebrew, configuring the merge driver, and executing a merge that yields a clean result with weave active.