# first-murex-mcp

Murex Enterprise MCP · RBC

walk the paths, rule things out

## what

There was no Murex tooling for LLMs, so I built my own (Logic Lake). When MCP came out, I rebuilt on it.

It investigates like I do: follow the logical paths, rule things out, let the data say what's wrong.

SMEs across multiple teams use it daily. 400+ hours per dev per year saved. Investigations went from hours to seconds.

Private RBC work, so I can't share specifics.

## why this approach

Murex is too big for one question to find what's wrong. You get a guess that sounds right.

Tools that interact fixed that. Different paths into the system, taken at once. When they agree, that's the answer.

## what breaks

Murex keeps evolving and the MCP has to keep up.

Every new team finds a corner it doesn't handle yet. Adoption moves faster than I can absorb the feedback.

## what i learned

Training an agent to think like a senior practitioner is more work than training it to answer fast. It's also the work that matters.

A working product doesn't sell itself. Demos, docs, presentation: people adopt what they like to see.

## lessons

• -a working product doesn't sell itself. people adopt what they like to see.
• -train the agent the way you think. logic paths first, then data, then the cause.
• -complexity is your problem, not theirs. if setup takes more than five minutes, adoption never happens.