code mode: better LLM UX on runtime added extensions #6188
Description
Please explain the motivation behind the feature request.
Right now, in code mode an inline MCP server gets a pretty bad name. I don't think it is easy for the LLM to guess it.
$ cargo run -p goose-cli --release -- run --text "Search for list_directory and read_text_file tools. Use them to list /tmp and read the first .txt file." --with-builtin code_execution --with-extension "npx -y @modelcontextprotocol/server-filesystem /tmp"
Finished `release` profile [optimized] target(s) in 0.25s
Running `target/release/goose run --text 'Search for list_directory and read_text_file tools. Use them to list /tmp and read the first .txt file.' --with-builtin code_execution --with-extension 'npx -y @modelcontextprotocol/server-filesystem /tmp'`
starting session | provider: tetrate model: gpt-5-nano
session id: 20251219_11
working directory: /Users/codefromthecrypt/oss/goose
─── search_modules | code_execution ──────────────────────────
terms: list_directory, read_text_file
─── read_module | code_execution ──────────────────────────
module_path: tmp_9zspqbtd/list_directory
─── read_module | code_execution ──────────────────────────
module_path: tmp_9zspqbtd/read_text_file
--snip--Describe the solution you'd like
I wonder if when there is no conflict, we can use the server name or parse some part of it. Probably other projects have a similar concern we can study.
- I have verified this does not duplicate an existing feature request