The lineage-plus-issues combination is what this has that a dashboard does not: an issue becomes traceable end to end rather than a row in a list. `get_table_dimension_coverage` answers the harder question, which is what you are not monitoring at all.
An MCP server for the Bigeye data observability platform, shipped as a Docker image. It reaches most of the platform: issues and incidents, metrics and profiling, the data catalog, lineage in both directions, dimension coverage, tags and the business glossary.
- Issues — `list_issues`, `get_issue`, `search_issues` by display number, `list_table_issues`, `list_related_issues` via lineage, `update_issue`, `create_incident` by merging related issues, `get_resolution_steps`
- get_current_user returns your id, which pairs with `list_issues(assignee_ids=[id])` to find what is assigned to you
- Metrics and profiling — `list_table_metrics`, `create_metric` with validation and column-type compatibility checks, `get_table_profile`, `create_profile_job`, `get_profile_job_status`
- Lineage — `get_lineage_graph` upstream, downstream or both, `search_lineage_nodes` by path pattern, `get_lineage_node`, `list_lineage_node_issues`
- Root cause and impact — `get_upstream_root_causes`, `get_downstream_impact`, `get_issue_lineage_trace`, `list_report_upstream_issues` for a BI report or dashboard
- Catalog — `search_schemas`, `search_tables`, `search_columns`, `list_data_sources`
- Sensitive data — `list_data_classes`, `get_scan_findings` showing where classified data was detected
- Dimensions, tags and glossary — including `get_table_dimension_coverage` for answering "what monitoring is missing?"
- Agent lineage tracking — record the data an agent touched with `lineage_track_data_access`, then `lineage_commit_agent` to persist it into the lineage graph
Docker, plus `BIGEYE_API_KEY` generated under Settings > API Keys, `BIGEYE_BASE_URL` for your instance, and `BIGEYE_WORKSPACE_ID`, which is the number after `/w/` in your Bigeye URL. `BIGEYE_DEBUG` turns on verbose logging and `BIGEYE_TELEMETRY=false` disables anonymous usage analytics — which record tool name, duration and success only, never arguments or results. The long-lived container path uses `mcp-wrapper.sh` with docker compose and expects the `.env` file beside `docker-compose.yml`; an ephemeral container per session is the alternative. Local development without Docker needs Python 3.12+.
Run a container — pull the image and point your client at it