On this page:

MCP server

The TriplyDB MCP server is what an AI assistant talks to when you connect it to TriplyDB. This page lists which revision of the protocol it speaks, and the tools it offers: the actions an AI assistant can take in TriplyDB on your behalf. Connect an AI assistant explains how to connect one.

Protocol

This server targets revision 2026-07-28 of the MCP specification. A client that speaks revision 2025-11-25 or earlier is served that instead.

A client identifies itself with a Client ID Metadata Document: an HTTPS URL where it publishes its own details. A client that can only register itself with Dynamic Client Registration cannot connect.

Tools

Your assistant may ask your permission before it uses a tool, and show you the tool's name when it does. The table below tells you what each tool does, so you know what you are approving. It also shows what an assistant can and cannot do in TriplyDB, which helps when a conversation doesn't go as expected.

Requests for anything the list doesn't cover go to support@triply.cc.

Tool What it does
get_current_user Reports who the connection is authenticated as, which groups you belong to, and which scope (read or write) it holds.
list Enumerates datasets, saved queries, stories, assets, graphs, groups, or group members, with pagination.
delete Permanently deletes a dataset, saved query, story, named graph, asset, service, group, or flow. Destructive and irreversible.
resolve_url Turns a console URL (a dataset, saved query, or story link) into the resource behind it.
update_account Edits a user's or a group's display name, description, or access level, or renames/moves it.
create_group Creates a group — including a nested group — that can own datasets, queries and stories on its members' behalf.
manage_group_member Adds, re-roles, or removes a group's members.
get_dataset Fetches a dataset's metadata, or a schema view — its classes, predicates, datatypes and example instances — for writing SPARQL against a dataset you haven't seen before.
create_dataset Creates a new, empty dataset under an account.
update_dataset Edits a dataset's metadata (title, description, tags, license, access level, example resources) and its prefix declarations.
import_from_urls Imports RDF into a dataset from one or more URLs. Runs in the background.
execute_sparql Runs a SELECT, CONSTRUCT, DESCRIBE, or ASK query against any dataset.
preview_sparql_update Dry-runs a SPARQL Update and reports the added/removed statement counts per graph, without committing anything.
execute_sparql_update Runs a SPARQL Update and commits it. Always preceded by a preview of the same update, so the diff can be confirmed first.
manage_service Creates a SPARQL or Elasticsearch service for a dataset, or syncs one with the dataset's current data.
create_query Saves a SPARQL query (with any visualization) as a named, shareable, versioned saved query.
update_query Edits a saved query's metadata (title, description, tags, access level), or appends a new version with new SPARQL or visualization.
read_visualization_guide Explains which query form and variable names each saved-query visualization needs, and how to configure it. Assistants read it before saving a query with any visualization other than Table.
create_story Publishes a data story: markdown paragraphs interleaved with embedded saved-query results.
update_story Edits a story's title, tags or access level, or replaces its content.

Flow toolsPreview

These tools are offered only on instances where flows are switched on.

Tool What it does
read_flow_guide Explains how a flow is written: how tasks, edges and variables fit together, and what each task type takes. Assistants read it before creating or changing a flow.
get_flow Reads a flow: its tasks, edges, variables and schedules, whether it is ready to run, and what keeps it from running.
create_flow Creates a flow. It is saved even when it is not ready to run yet, and the result lists what keeps it from running.
update_flow Replaces a flow with a changed version of it. Every save keeps the earlier version.
get_flow_revisions Lists the saved versions of a flow, or reads one, so an earlier version can be restored.
run_flow Starts a run of a flow, with a value for each of its variables. The run acts with your rights, so its tasks can create, overwrite and delete datasets and other resources.
get_flow_runs Lists a flow's runs, or reads one: what started it, its status, why it failed, and the state of each task.
manage_flow_schedule Creates, changes or deletes a schedule that runs a flow at set times. A scheduled run acts with the rights of the schedule's owner.