refactor: drop jupyter-compute from the Jupyter MCP integration

[KadenMc]

Summary

aexp install --with-jupyter wired two near-duplicate Jupyter MCP servers into a consumer's .mcp.json:

• jupyter — uvx jupyter-mcp-server, MCP_SERVER mode, target URL+token supplied at runtime via connect_to_jupyter
• jupyter-compute — npx mcp-remote proxy to a fixed cluster /mcp endpoint, JUPYTER_SERVER mode

jupyter-compute could not retarget to a different node without a .mcp.json edit + MCP restart — which breaks the multi-node workflow the /aexp-jupyter-connect / /aexp-jupyter-discover commands are built around. It was otherwise a near-duplicate of jupyter and a standing "which server do I use?" confusion surface (the old docs even mis-steered toward it).

This reduces the integration to a single server, jupyter.

Changes

• install.py — _jupyter_mcp_entries() emits only the jupyter entry; the additive .mcp.json merge no longer special-cases jupyter-compute. Docstrings updated.
• cli.py — --with-jupyter help text.
• Slash commands /aexp-jupyter-iterate and /aexp-promote-nb, AGENTS.md, docs/setup/jupyter-mcp.md, README.md — retargeted to the single-server mcp__jupyter__* tool family. The mis-steering "jupyter-compute is functionally a superset" troubleshooting line and the jupyter-compute-specific investigation-log rounds are removed.
• tests/test_install.py — 5 jupyter tests updated.
• CHANGELOG.md — [Unreleased] entry.

Lost capability

The two jupyter-mcp-tools UI-delegated tools. notebook_run-all-cells was already 404-broken upstream, so that costs nothing; notebook_get-selected-cell ("which cell is the user looking at") is genuinely gone — the affected slash commands now ask the user for the notebook/cell or cross-check aexp.jupyter.init().attached_notebooks.

Scope notes

• The cluster-side [jupyter] pip extra and the aexp jupyter setup extension recipe are intentionally unchanged — trimming them risks a working cluster deployment and is a separate migration.
• A consumer .mcp.json written by an earlier --with-jupyter install keeps its jupyter-compute entry (the merge is additive-only and never deletes servers); remove it by hand for the cleanup.

Testing

• Full suite: 454 passed, 5 skipped (pre-existing environmental skips), exit 0
• ruff check clean on changed sources
• mypy shows only pre-existing errors on lines untouched by this change

🤖 Generated with Claude Code