As ClickHouse is growing in prominence, businesses such as banks, hedge funds, exchanges, brokers and trading tech vendors are increasingly interested in using it to store and analyse their data. It is for instance increasingly being baked off against kdb+ as a more open and

A lot of teams find that their applications backed by MySQL and PostgreSQL slow down over time, particularly any embedded analytics, reports and charts. Resolving this past a certain scale can be tricky. If you do it in the application you might be into pre-computing numbers of

Everyone talks about the performance of ClickHouse, but there's also an important simplification business case too. I have worked in data for some time now, and I have to say a lot of the data engineering work moving data between systems, transforming it with Spark, setting up

See below for a cool demo of a sports analytics scenario I've built with ClickHouse. The frontend was built with Claude Code in a few hours on top of some existing simulation code I had. It consists of a stream of player data which is hitting ClickHouse, with a position for

When LLMs first hit the scene I had a play with "generative BI" whereby the LLM would create SQL. Early attempts were pretty fragile to say the least, but nowadays with MCP, better models and an agentic experience powered by Claude it is really getting close to prrime time.

I think the statement that ClickHouse can't do joins is now officially out of date. It smoked Snowflake and Databricks here. The video is worth watching too. clickhouse.com/blog/join-me-i…

Experimenting with connecting both Claude and Gemini to ClickHouse via the MCP server. I threw some data analysis type tasks at it (analyse a property porice dataset.). Overall, Claude seems to be ahead for generative BI, deeper analysis, more accurate, faster.

Connecting Gemini CLI or Claude Code to your back end database via an MCP server is a really remarkable development experience. Here, I've asked it to build a low code app and just given it a hint at the database table. Gemini CLI logs into the database, locates the relevant

I used both Gemini and Claude to query ClickHouse using natural language. I used a simple SQL test found online to try to compare them objectively.  Results were interesting, but I will say I think being able to do this is really becoming viable nowadays.

Working on a new TCO calculator to properly bake off the costs of ClickHouse Cloud vs self managed open source. People think that open source is cheaper, but by the time you've accounted for your infrastructure, the fact that you can't scale up and down so easily, and the fact

Traditionally, ClickHouse hasn't been that fast when it comes to joins. However, in recent releases it is now performing 20x faster in a TPC-H benchmark. Using another public benchmark a set of join heavy queries over a 500m row dataset now smokes Databricks and Snowflake. I

There is so much potential in using AI to analyse data now text to SQL is ready for prime time. Here I analysed a dataset in plain English. I then built a low code app for exploration directly in Claude desktop. Next, I asked Claude to complete some unstructured internet

I've been working on a crypto project recently and found ClickHouse to be such a great fit. Whilst the Bitcoin block time was 10 seconds, many of the newer chains create blocks at < 500ms with thousands of transactions and events which most databases will struggle with.

Remember when we talked about ClickHouse and MCP? It's like peanut butter and jelly for data and AI! 🥜🍇 clickhou.se/4na2eej Curious how ClickHouse fits into the AI revolution? Our blog post explores: 🔸 What MCP is and why it matters 🔸 How ClickHouse accidentally became