#Everything Wrong with MCP
1 messages · Page 1 of 1 (latest)
could you please detail that part 🙏
Esp when it comes to some of the issues Ive had with my own QA agent
They asked “how many times have I said ‘AI’ in docs I’ve written” and after around 30 read_file(…) operations the agent gives up as it nears its full context window. It returns the count among only those 30 files which the user knows is obviously wrong. The MCP’s set of tools effectively made this simple query impossible. This gets even more difficult when users expect more complex joins across MCP servers, such as: “In the last few weekly job listings spreadsheets, which candidates have ‘java’ on their linkedin profiles”.
I keep running into problems that looks just like this one.
My QA agent generates QA reports for dagger modules. It does a good job at that.
But then when I feed it a repo full of reports and ask it to generate an index.html with a table, links, status to the other modules its falls apart. After reading this I suspect it may be due to the context window being full or something along those lines 🙂