Get the code execution tools

Returns run_r_code and run_bash. Both run model-written code with your user account's access. Each call runs in a separate process, which is not a sandbox. permissions_standard() denies both tools; allow them with r_code = TRUE and bash = TRUE in Permissions(). For an OS sandbox, use tools_mcp_repl().

The processes don't inherit your environment variables, and R there doesn't read .Renviron. They get the variables that locate programs, libraries, locales and temporary files, such as PATH, HOME, LANG and TMPDIR, plus the ones you name in env. This keeps keys out of what the code is given, not out of its reach: code running as your user account can still read your R session's starting environment through the operating system, and any file your account can read, .Renviron included. To keep keys from the code, run it under an account that can't read them, or in a sandbox.

Usage

tools_code(env = NULL)

Arguments

env

The names of other environment variables the code may read, such as c("HTTPS_PROXY", "NO_PROXY") behind a proxy. Proxy settings match in either case, so "HTTPS_PROXY" passes https_proxy too. "inherit" passes your whole environment, including every key it holds.

Value

A list of tools.

See also

tool_run_r_code, tool_run_bash

Examples

if (FALSE) { # \dontrun{
agent <- Agent$new(
  chat = ellmer::chat("openai/gpt-6-luna"),
  tools = tools_code(),
  permissions = Permissions(r_code = TRUE, bash = TRUE)
)
} # }