abba-mcp
Official
# abba-mcp: ABBA for AI assistants (experimental)
Lets an AI assistant drive [ABBA](https://abba-documentation.readthedocs.io) (Aligning Big Brains and Atlases) to
register brain sections to an atlas. The assistant runs ABBA's commands, looks at what they produced, and once the
result is right saves the session (`.abba` state file) and a Groovy script that replays every step.
You can follow along and take over at any time: by default, ABBA's window opens on your screen as usual.
It is an [MCP](https://modelcontextprotocol.io) server, so it is not tied to one assistant: it works with any
application (MCP client) that can start a local MCP server, and any model it offers that calls tools and reads
images (the assistant looks at pictures of your slices).
## Install
Follow the [installation page](https://abba-documentation.readthedocs.io/en/stable/installation/installation.html#install-assistant)
of the ABBA documentation: VS Code (one click), Claude Desktop and Claude Code, Codex, and any other MCP client;
the first run, running without a window, privacy, and what to do if something goes wrong.
In short: install [uv](https://docs.astral.sh/uv/getting-started/installation/), then register the command
`uvx abba-mcp` as a local (stdio) MCP server in your application, or click:
[](https://vscode.dev/redirect/mcp/install?name=abba&config=%7B%22command%22%3A%22uvx%22%2C%22args%22%3A%5B%22abba-mcp%22%5D%7D)
The version is the ABBA version it runs (`0.25.0`), with `.post1`, `.post2`... for a fix of this server alone.
Questions about ABBA itself: the [ABBA documentation](https://abba-documentation.readthedocs.io/en/v0.25.0/) of this
version and the [image.sc forum](https://forum.image.sc/tag/abba).
Developing or releasing this bundle: see [DEVELOPMENT.md](DEVELOPMENT.md).
TDQS
Scored across 16 tools
Each tool targets a distinct resource+action: search/describe/run/status for modules, read/respond/close for dialogs, picture/view for windows, search/read for docs. The descriptions even explicitly cross-reference each other (e.g. use fiji_module_status to follow a long fiji_module_run), leaving little room for misselection, aside from a mild adjacency between the three script-related tools that the text clearly disambiguates.
A predictable namespace_noun_verb, snake_case pattern dominates (fiji_module_search, fiji_ui_dialog_respond, abba_docs_read, abba_window_view) with clear prefixes (fiji_module_, fiji_ui_, abba_). Minor deviations exist: fiji_module_script has no verb, and fiji_module_run_script/abba_window_picture break the verb-final rhythm, but the scheme is still highly readable.
16 tools is at the heavier end but each covers a genuine, non-redundant capability (module lifecycle, session replay, UI introspection/interaction, window control, docs, stop). Nothing feels padded, though the module/script/status cluster could conceivably be tightened.
The surface covers discovery, execution, monitoring, scripting, live-object inspection, UI read/interaction, window rendering, and docs — a full lifecycle for driving Fiji/ABBA. Minor asymmetries remain, e.g. an explicit stop but no explicit start/restart (start is implicit in run), and no standalone session save/load beyond script replay, but agents can work around these.