Sidecar

Search for data assets using identifier matching, keyword search, semantic similarity, or hybrid fusion. Use 'identifier' mode (default) when you know the asset name or a pattern (supports * wildcards). Use 'keyword' when searching by domain terms or metadata keywords. Use 'semantic' for natural language queries about what the data represents. Use 'hybrid' to combine all approaches for the best recall. Returns a ranked list of matching assets with identifiers, types, and relevance scores.

Find everything downstream of a data asset — tables, dashboards, and reports that would break or be affected if this asset changes. Call this automatically after the user modifies any SQL file, dbt model, or source definition, or whenever they ask 'what depends on X?' or 'what breaks if I change Y?' Returns a breakdown of affected assets by tool type and a full list of downstream nodes.

Fetch detailed information about a data asset by its identifier. Resolves the asset, then fetches the requested sections in parallel. Use this after advanced_search or resolve_asset_id to get deeper information about a specific asset. Sections: - overview: asset description, importance, owner, metadata - columns: full column schema with types and descriptions - tests: dbt test coverage (only for assets with a dbt parent) - context_counts: cheap count of linked tickets and Slack threads - tickets: full Jira/Linear ticket content - slack: full Slack thread content (truncated to first 5 messages per thread)

Get a summary of everything available in this customer's Sidecar catalog. Returns the connected tools (e.g. SNOWFLAKE, DBT_CLOUD, LOOKER), asset type breakdown with counts, all available tag names, and a customer-provided company summary describing the company's domain and data landscape (if set). Call this at the start of any open-ended exploration (e.g. 'what data do we have?', 'show me all Snowflake tables'), or when a search returns no results and you need to understand what's actually in the catalog before trying again. Use the customer_summary field (when present) to understand the business context behind the data.

Trace column-level lineage — which upstream columns feed into a column and which downstream columns consume it. Use this when the user asks 'where does this column come from?', 'what columns feed into X.col?', or needs to understand column-level data flow. Returns the target column, upstream columns (with levels), and downstream columns (with levels).

Return the clean OSI semantic contract currently deployed to this MCP token's Context Garage surface.

Fetch the full text content of a file from the customer's knowledge base. Use this after calling list_knowledge_base_files to get file IDs. Works with any file in the knowledge base — playbooks, metric definitions, nuances, or captured context summaries.

Trace the full data lineage for an asset — both upstream sources and downstream consumers. Use this when the user asks 'where does this data come from?', 'what feeds into X?', or needs the raw lineage graph. For pure downstream impact (blast radius), prefer `calculate_downstream_impact` instead.

List all files in the customer's knowledge base. The knowledge base contains company-specific reference material — playbooks, metric definitions, business nuances, and conventions — uploaded by the customer's team. Returns file names, paths, types, and IDs. Pass a file ID to `get_knowledge_base_file_content` to read its content. Use the optional folder_path parameter to narrow results to a specific folder.

Look up any data asset by name across the full catalog (dbt, Snowflake, BigQuery, Looker, etc.). Returns the asset's columns, description, owner, tags, and its **Sidecar asset identifier** (`asset_identifier` field) — the fully-qualified internal name used by lineage and impact tools (e.g. `model.jaffle_shop.fct_orders`, `PROD_DB.ANALYTICS.FCT_ORDERS`). When a single asset is resolved, the response also includes a `context_summary` with counts of linked tickets (Jira/Linear) and Slack threads. If those counts are non-zero and relevant to the user's question, follow up with `describe_asset` using include=['tickets'] or include=['slack'] to fetch the actual content. Call this first whenever the user refers to a specific table, model, view, or dashboard.