Connect Claude to ClickHouse

Explore a ClickHouse server's databases, tables and columns, then run read-only SQL against it. Toolspoke puts 9 of its actions behind one MCP endpoint that Claude, Cursor, and Codex all speak.

Connect View pricing

Connection: Provider REST API
Authentication: User and password
Actions exposed: 9
Cost per call (typical): 1 credit
Adapter: Maintained by Toolspoke

Connected in three steps

1
Install ClickHouse
Open the marketplace in your workspace, add ClickHouse to the project your agents work in, and it appears on the gateway immediately.
2
Connect the credential
Authenticate with user and password. Where to get one, and what it has to be able to reach, is the next section.
3
Point your agent at the gateway
Give your client one address, https://toolspoke.com/mcp. Claude Code takes it as a command, Claude and Claude Desktop add it as a custom connector, and Cursor, Codex and VS Code each read it from a config file of their own.

.mcp.json

{
  "mcpServers": {
    "toolspoke": {
      "type": "http",
      "url": "https://toolspoke.com/mcp"
    }
  }
}

One block covers every tool you have installed. ClickHouse shows up in the client as soon as your policy allows it, and so does everything else you install later.

Where the address goes, per client

Claude Code: Run it in your project, then /mcp to sign in
claude mcp add --transport http toolspoke https://toolspoke.com/mcp
Claude and Claude Desktop: Settings, then Connectors, then Add custom connector
https://toolspoke.com/mcp
Cursor: ~/.cursor/mcp.json, or .cursor/mcp.json for one project
{ "mcpServers": { "toolspoke": { "url": "https://toolspoke.com/mcp" } } }
Codex: ~/.codex/config.toml
[mcp_servers.toolspoke] url = "https://toolspoke.com/mcp"
VS Code: .vscode/mcp.json, or the MCP: Add Server command
{ "servers": { "toolspoke": { "type": "http", "url": "https://toolspoke.com/mcp" } } }

What ClickHouse asks for

User and password. You provide it once, when you install the connector. Toolspoke encrypts it at rest and decrypts it only for the length of a single call, and the gateway attaches it to the outbound request itself, so it is never part of the arguments an agent sends.

HTTP endpointRequired: The ClickHouse HTTP interface root, with its port. ClickHouse Cloud serves 8443 over TLS; a self-managed server is usually 8123. No path. The native protocol ports (9000, 9440) are not HTTP and will not work.; https://abc123.eu-central-1.aws.clickhouse.cloud:8443
User and passwordRequired: The database user and its password joined by a colon, sent as HTTP Basic. Give that user SELECT and nothing else: this connector cannot write, but the grant is the real boundary and the read-only setting is the second line of defence.; default:my-password

What Claude can do in ClickHouse

9 actions, each one declared and named by the connector rather than discovered at runtime. A workspace policy grants a person all of them, a hand-picked selection, everything on the read side, everything on the write side, or none.

Reads: 9Reads
Writes: 0Writes
Destructive: 0Destructive

Reads

Fetches data and changes nothing.

server_info
Report the ClickHouse version, the connected user, the default database, the server timezone and how long the server has been up. Call this first to confirm the connection and the credential work.
list_databases
List the databases on this server with their engine and comment. Use it to find where the application's tables live before calling list_tables.
list_tables
List tables and views with their engine, row count, size on disk, sorting key and comment. The database and table filters are SQL LIKE patterns, so % matches everything and events% matches a prefix. total_rows and total_bytes come back null for engines that cannot report them, such as views.
describe_table
Describe one table: every column with its position, type, default expression, compression codec, key membership and comment. Use this before writing a query against an unfamiliar table.
get_create_statement
Return the CREATE TABLE statement for one table, which is the only place the engine settings, TTL rules and codecs appear together. Use it when describe_table does not explain a table's behaviour.
table_sizes
List the largest tables by bytes on disk, counted from the active parts only, with compressed and uncompressed totals and a part count. Use it to find what is filling the server.
running_queries
List the queries running on this server right now, longest first, with the user, elapsed time, rows read and memory in use. Read-only - it reports, it cannot cancel anything. On ClickHouse Cloud this is one node's view, not the whole service.
query_history
List finished queries from system.query_log over the last N minutes, slowest first, with duration, rows read, memory and any exception. Use it to find what is slow or failing. The log is flushed every few seconds, so a query run moments ago may not be there yet, and on ClickHouse Cloud this is one node's traffic rather than the whole service.
run_query
Run one read-only SQL statement and return the rows with their column types. The server is put in readonly mode for the request, so INSERT, ALTER, CREATE, DROP and SET are refused and a SETTINGS clause in your own SQL cannot turn that off. There is a 30 second execution limit, and a result larger than `max_rows` is refused rather than truncated - so put your own LIMIT in the statement. EXPLAIN works here too: send the whole EXPLAIN statement as the sql.

What it will not do

Enforced by the gateway rather than left to convention, which is why each of these can be stated flatly.

It cannot call anything else: The 9 actions above are the whole of it. A call to any other name is refused before it reaches ClickHouse rather than forwarded on, and connecting your account does not add to the list: it is fixed by the connector, not discovered at run time.
It only reads: Every action here reads. Nothing this connector can do changes anything in ClickHouse.
It reaches no further than your credential: Toolspoke holds no access to ClickHouse of its own. Every call carries the credential you stored and nothing besides, so whatever that credential cannot reach, this connector cannot reach either.
It never hears from ClickHouse: Nothing is pushed to it. There is no webhook, no subscription and no polling, so this connector cannot notice by itself that something changed in ClickHouse. An agent has to ask.
It does not smooth over provider limits: Toolspoke does not retry, queue or back off around ClickHouse's own rate limits. A call that ClickHouse refuses comes back to the agent as a failed call.

Before you connect it

What can Claude do in ClickHouse?

9 named actions: 9 that only read. They include server_info, list_databases and list_tables. Nothing outside that list is reachable: the connector declares each operation by name rather than proxying whatever an agent asks for.

What credentials does the ClickHouse connector need?

User and password. The connector asks for http endpoint and user and password. Values are encrypted at rest and attached to the outbound request by the gateway, so they are never part of the arguments an agent sends and never reach the audit log.

Does the ClickHouse connector work with Cursor and Codex, or only Claude?

Any client that speaks MCP, and every one of them gets the same 9 actions. There is a single address, https://toolspoke.com/mcp. Claude Code adds it with claude mcp add --transport http, Claude and Claude Desktop take it as a custom connector in settings, Cursor reads it from .cursor/mcp.json, Codex from ~/.codex/config.toml, and VS Code from .vscode/mcp.json. Each of them signs in to the gateway itself, so there is no key to paste.

What does the ClickHouse connector not do?

The 9 actions above are the whole of it. A call to any other name is refused before it reaches ClickHouse rather than forwarded on, and connecting your account does not add to the list: it is fixed by the connector, not discovered at run time. Every action here reads. Nothing this connector can do changes anything in ClickHouse. Toolspoke holds no access to ClickHouse of its own. Every call carries the credential you stored and nothing besides, so whatever that credential cannot reach, this connector cannot reach either. Nothing is pushed to it. There is no webhook, no subscription and no polling, so this connector cannot notice by itself that something changed in ClickHouse. An agent has to ask. Toolspoke does not retry, queue or back off around ClickHouse's own rate limits. A call that ClickHouse refuses comes back to the agent as a failed call.

Can I limit which actions an agent can call?

Yes, in two places. The project switches ClickHouse's actions on and off one at a time, for everyone in the project at once, and the screen groups them by read, write and destructive so turning off everything that deletes is one click. An individual agent key can then be narrowed further, to particular toolkits in a project and to particular actions in a toolkit. Whatever it was granted, a key never reaches a project its owner cannot.

What gets recorded when an agent calls ClickHouse?

Every attempt, with the agent that made it and the person that agent belongs to, the full request payload, the response payload, the status, the duration, and the credits spent. Values whose key names a secret are masked out before the record is shown to anyone. An operation the connector marks as not retained never has its response body written at all, so the gateway keeps no second copy of what was read.

More database connectors

Browse all of them