martypitt

I was doing some client work recently at a bank, where most of their engineering is offshored one of the big offshore companies.

The offshore team had to access everything via virtual desktops, and one of the restrictions was no virtualisation within the virtual desktop - so tooling like Docker was banned.

I was really surprsied to see modern JVM development going on, without access to things like TestContainers, LocalStack, or Docker at all.

To compound matters, they had a single shared dev env, (for cost reasons), so the team were constantly breaking each others stuff.

How common is this? Also, curious what kinds of workarounds people are using?

Hey HN

There's been a lot of momentum this year around the MCP ecosystem - like yesterday’s announcement from Anthropic enabling MCP access via Claude's web interface.

I'm curious what kind of enterprise adoption is actually happening. Are companies putting real budget into MCP-focused tooling yet?

For context: I'm a founder in the data federation and integration space. Our product is used by developers, so MCP tooling is something we're naturally evaluating. But among our (mostly banking) customers, I’m not seeing signs of adoption yet - it's not even on the radar.

I've spoken with several teams that use our platform: none are currently exploring MCP initiatives or LLMs in a structured way, beyond widespread use of tools like GitHub Copilot. (But, these are all enterprise customers, so it's possible evaluation is happening elsewhere in the company).

Would love to hear what others are seeing - especially if you’re building in this space or selling to enterprises.