Model Context Protocol

Tentris provides an endpoint for the Model Context Protocol (MCP). AI agents, such as Claude Code, use it to explore and query your data.

⚠️ Warning: The MCP endpoint is currently an experimental feature.

Enabling the Endpoint

Add the [serve.mcp] section to the configuration file and restart Tentris.

[serve.mcp]

The endpoint is then available at http://<host>:<port>/mcp. Without this section, /mcp does not exist.

What Agents Can Do

The endpoint offers three tools:

• dataset_statistics: the size of the dataset and of its largest named graphs.
• describe_schema: the most common classes and predicates of the dataset.
• sparql_query: runs read-only SPARQL queries (SELECT, ASK, CONSTRUCT and DESCRIBE).

All tools are read-only. Tool calls have the same time limit as SPARQL queries (limit-query-duration-secs).

Authentication

If auth is disabled, the endpoint is open to everyone, like all other endpoints.

If auth is enabled, the endpoint accepts only a bearer token:

Authorization: Bearer tnt_...

The owner of the token must be in the query group. The endpoint does not accept passwords, cookies or guests, even with allow-guests = true. To make a token, refer to Tokens.

⚠️ Warning: Use HTTPS ([serve.tls]) when the server is not on localhost. Otherwise, the token is sent in plain text.

Connecting Claude Code

claude mcp add --transport http tentris http://localhost:9080/mcp \
    --header "Authorization: Bearer tnt_..."

Replace tnt_... with your token. Without auth, leave out the --header line.

To check the connection, run claude mcp list. The output should contain this line:

tentris: http://localhost:9080/mcp (HTTP) - ✔ Connected

You can now ask Claude questions about your data.

What can you tell me about the data stored in Tentris?

Note: By default, claude mcp add registers Tentris only for the current project. Use --scope user to register it for all your projects.

To remove Tentris from Claude Code, run this command:

claude mcp remove tentris