Skip to content

MCP tools

Every tool call is authenticated, authorized by source roles, shaped by policy, redacted and audited, exactly like a query on any other protocol. Tools named agentgate_* are deprecated aliases of the interlock_* tools, kept until 1.2.0.

List the actions the caller’s source roles allow and deny on each granted data source, and for SQL sources the tables and columns the caller may read, with any table a policy would refuse marked. Use this to plan before querying rather than discovering limits by being refused. Policy rules may still restrict an individual request, so an allowed action here is permission to attempt, not a guarantee of success.

Argument Type Required Meaning
source_id string no Limit the answer to one source. Omit for every granted source.

Get schema information for a data source

Argument Type Required Meaning
source_id string yes Data source identifier

Search the discovery index for relevant data assets

Argument Type Required Meaning
audience string no Optional audience filter
category_path string no Optional hierarchical category path
doc_types array no Optional document type filters
entities array no Optional entity filters
include_related boolean no Whether to include related documents
limit integer no Maximum number of results (default 10)
query string yes Natural language search query
source_id string yes Required data source scope
source_types array no Optional source type filters
time_range object no Optional time range filter

List registered data sources visible to the caller

Execute a SQL query against a registered data source

Argument Type Required Meaning
source_id string yes Data source identifier
sql string yes SQL query to execute

Find documents related to a discovery asset within the same authorized source

Argument Type Required Meaning
asset_id integer yes Discovery asset identifier
limit integer no Maximum number of related documents (default 10)