We open-sourced , the Bash execution engine used by to reduce our token usage, improve the accuracy of the agent's responses, and improve the agent's overall performance.bash-toolour text-to-SQL agent that we recently re-architected gives your agent a way to find the right context by running bash-like commands over files, then returning only the results of those tool calls to the model.bash-tool Context windows can fill up quickly if you include large amounts of text into a prompt. As agents tend to do well with Unix-style workflows like , , , and pipes, with you can now keep large context in a filesystem, and let the agent use those commands to retrieve smaller slices of context on demand.findgrepjqbash-tool local, provides , , and tools for agents, working with both in-memory and sandboxed environments, and:bash-toolbashreadFilewriteFileAI SDK If you need a real shell, a real filesystem, or custom binaries, you can run the same tool against a Sandbox-compatible API for full VM isolation.

Read more runs on top of , which interprets bash scripts directly in TypeScript without a shell process or arbitrary binary executionjust-bash you can preload that filesystem with your files at…