MCP Server Tools Reference

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/neo4j-labs/neocarta/llms.txt

Use this file to discover all available pages before exploring further.

The Neocarta MCP server exposes two categories of tools to AI agents: catalog tools that are always available from schema data alone, and search tools that are registered conditionally based on the indexes present in Neo4j at startup. Every search tool returns a TableContext record — a table together with its columns, types, example values, and foreign-key references — giving agents everything they need to write a correct query.

​

Catalog Tools

Catalog tools are always registered, regardless of whether embeddings or search indexes exist. They provide structural orientation for an agent navigating an unfamiliar database.

---

​

list_schemas

Lists all schemas in the semantic graph alongside the database each belongs to.

• Parameters: none
• Returns: list of { schema_name, database_name }
• Use when: an agent needs to orient itself before querying — a good first call when the schema structure is unknown.

---

​

list_tables_by_schema

Lists every table within a named schema.

• Returns: list of { schema_name, table_names[] }
• Use when: the agent knows which schema is relevant and wants to enumerate available tables before deciding which to retrieve context for.

---

​

get_full_metadata_schema

Returns the complete metadata record for every table in the database — all columns, types, nullability, primary/foreign key flags, example values, and REFERENCES.

• Parameters: none
• Returns: list of TableContext for every table, ordered by table name

---

​

Search Tools

Search tools are registered conditionally at server startup based on which indexes exist in Neo4j. For each searchable label (Table, Column) the server registers only the single highest-priority tool whose prerequisites are met. See Search Strategy Auto-Detection for the full priority order. All search tools share the following common parameters:

---

​

Vector Search Tools

Requires: a VECTOR index on the target label. The query text_content is embedded at call time using the configured EMBEDDING_MODEL and compared against stored node embeddings via cosine similarity.

​

get_context_by_table_vector_search

Anchors on the most semantically similar table nodes, then expands each to its full column set.

• Default max_tables: 10
• Retrieval: queries table_vector_index, score threshold > 0.5
• Use when: the query describes a general concept or entity (e.g. "customers", "sales transactions").

---

​

get_context_by_column_vector_search

Anchors on the most semantically similar column nodes, groups them by parent table, and returns each table with only the matched columns.

• Default max_tables: 5
• Retrieval: queries column_vector_index, score threshold > 0.5, then groups to parent tables by average anchor score
• Use when: the query references a specific field or attribute (e.g. "customer email", "order total").

---

​

get_context_by_schema_and_table_vector_search

Anchors first on the most relevant schemas, then filters to tables within those schemas whose embedding scores are near or better than the schema score.

• Default max_tables: 5
• Default search_top_k: 5 (applied to schema candidates)
• Retrieval: queries schema_vector_index (top 5, score > 0.5), then includes tables where table_score > schema_score - 0.2; results ordered by schema score DESC, table score DESC
• Use when: the query is broad and may span multiple schemas (e.g. "everything related to billing").

---

​

Full-Text Search Tools

Requires: a FULLTEXT index on the target label. Supports Lucene query syntax in text_content. No embeddings or API calls required at query time.

​

get_context_by_table_full_text_search

Anchors on table nodes by full-text matching on table name and description.

• Default max_tables: 10
• Retrieval: queries table_full_text_index, results ordered by Lucene score
• Use when: the query contains literal tokens expected to appear verbatim in table names or descriptions (e.g. "orders", "fct_revenue").

---

​

get_context_by_column_full_text_search

Anchors on column nodes by full-text matching on column name and description, then groups to parent tables by average anchor score. Returns only the matching columns per table.

• Default max_tables: 5
• Retrieval: queries column_full_text_index
• Use when: the query references specific column-name tokens (e.g. "customer_id", "total_amount").

---

​

Hybrid Search Tools

Requires: both a VECTOR and a FULLTEXT index on the target label. Runs both branches in parallel with the same text_content, min-max normalizes scores within each branch by the branch maximum, then merges per node by taking the stronger of the two scores.

​

get_context_by_table_hybrid_search

Hybrid vector + full-text search at the table level.

• Default max_tables: 5
• Use when: the query mixes conceptual phrasing with literal tokens you expect to appear in table metadata.

---

​

get_context_by_column_hybrid_search

Hybrid vector + full-text search at the column level, returning only matching columns per table.

• Default max_tables: 5
• Use when: the query references field-level concepts alongside literal column-name tokens.

---

​

Business-Term Hybrid Search Tools

Requires: a VECTOR index on the target label + a FULLTEXT index on the target label + businessterm_full_text_index + at least one :BusinessTerm node in the graph. These tools extend the hybrid search by routing the full-text branch through the business glossary. The full-text branch finds matching BusinessTerm nodes and then surfaces only the tables or columns that are also connected to those terms via TAGGED_WITH. This makes the tools effective for domain vocabulary that does not appear verbatim in table or column metadata.

​

get_context_by_table_business_term_hybrid_search

Hybrid search at the table level, with the full-text branch bridged through :BusinessTerm tags.

• Default max_tables: 5
• Retrieval: UNION of table_vector_index (vector branch) and businessterm_full_text_index + table_full_text_index where TAGGED_WITH (business-term branch); per-branch normalization, max-per-node merge
• Use when: the query uses business-glossary language (e.g. "average order value", "gross merchandise value") that is tagged to relevant tables.

---

​

get_context_by_column_business_term_hybrid_search

Hybrid search at the column level, with the full-text branch bridged through :BusinessTerm tags. Returns only matching columns per parent table.

• Default max_tables: 5
• Retrieval: UNION of column_vector_index (vector branch) and businessterm_full_text_index + column_full_text_index where TAGGED_WITH (business-term branch)
• Use when: business-glossary language maps onto field-level concepts via column tags (e.g. "customer acquisition cost" tagged to a cost column).

---

​

What Each Tool Returns

Every search tool returns a list of TableContext records. Each record contains: The REFERENCES field is what enables agents to discover join paths automatically:

Which customers placed the largest orders last quarter? The agent calls get_context_by_table_hybrid_search, finds orders and customers, sees orders.customer_id → customers.id in the REFERENCES, writes the JOIN, and executes the query.

​

Tool Registration Logic