| name | authoring-pipelines |
| description | Use when the user wants a bagel data pipeline — reducing a log to windows around events ("keep 30s around every hard brake"), recurring snippets/GIFs/exports, or any tasks-and-gates automation over robot data. |
Authoring bagel pipelines
The bagel MCP server owns the authoring workflow, including the
reduce-vs-snippet decision, window/debounce extraction, and the
preview-before-run rule. Do not write pipeline YAML from memory:
- Call
run_poml_capability with poml_path="./src/agent/compose/pipeline.poml".
- Follow it exactly. In particular: always call
preview_pipeline and show the
user the summary (events found, data kept) BEFORE running anything.
- Use
list_pipeline_capabilities for the exact task/gate module paths and
arguments — never guess them.
- Execute the approved config through the MCP tools (
run_pipeline, or
run_pipeline_batch for many sources). The capability also mentions a host
CLI (run.py); that path is for users at a terminal in the repo, not for
plugin sessions — do not shell out to it.
If the connection fails, the server container is not running — see
references/formats.md.