3 ms·

> The Problem > You: "Remember that auth bug we fixed?" > Claude: "I don't have memory of previous conversations." > You: "We spent 3 hours on it yesterday"

> The Problem > You: "Remember that auth bug we fixed?" > Claude: "I don't have memory of previous conversations." > You: "We spent 3 hours on it yesterday" > Claude: "I'd be happy to help debug from scratch!" People are twisting themselves into knots trying to solve issues like this. Let's use the mental model of "coding agents are not magic, treat them like humans." The boring old Jira MCP [0] completely solves the posed problem. > You: "Remember that auth bug we fixed?" > Claude: Called Atlassian Rovo > Claude: "Yes, I see PROJ-123 with commit SHAs, comments on what we decided and why. How would you like to proceed?" If all LLMs disappear, you still have human readable Jira tickets to continue working on your project, instead of something like QR codes in an MP4. [0] I am not trying to promote Jira or MCPs here, use whatever you want. I went back to Jira because its usage patterns are very well represented in the training data, and their MCP is not in beta.

slowmovintarget 6mo ago
Even simpler: wedow/ticket with in-repo tracking of tasks. No MCP, no Jira. (This is a markdown-only replacement for Beads.) But yes, wiping the context clean is no problem, in fact it's preferred hygiene.
- consumer451 6mo ago
  Thanks for suggesting this. Prior to going back to Jira, I had been curating my own set of .md docs & tickets in my repo, but it got way our of hand trying to roll my own. Given Atlassian's AI training news, I will be trying out tk.
lostmsu 6mo ago
You can also tell Claude/Codex/whatever to look up previous conversations in respective folders.
- consumer451 6mo ago
  Yes, I go even further. In-repo, I have a chats folder that my /done skill fills with ~"what we did, and didn't accomplish in this chat. Blah blah (a few more instructions) - finish with a great hand off to the next chat to continue the work." I run that anytime I approach 50% of the context window, as all models get dumb at that point. Then /clear, then /effort max just to be safe, then "please ingest chats/2026-01-01-00-00-what-we-did.md and proceed." It's a very purposeful custom /compress that works far better in my experience. If I ever hit auto-compress, I have failed as a Claude jockey.

slowmovintarget 6mo ago

Even simpler: wedow/ticket with in-repo tracking of tasks. No MCP, no Jira. (This is a markdown-only replacement for Beads.) But yes, wiping the context clean is no problem, in fact it's preferred hygiene.

consumer451 6mo ago
Thanks for suggesting this. Prior to going back to Jira, I had been curating my own set of .md docs & tickets in my repo, but it got way our of hand trying to roll my own. Given Atlassian's AI training news, I will be trying out tk.

lostmsu 6mo ago

You can also tell Claude/Codex/whatever to look up previous conversations in respective folders.

consumer451 6mo ago
Yes, I go even further. In-repo, I have a chats folder that my /done skill fills with ~"what we did, and didn't accomplish in this chat. Blah blah (a few more instructions) - finish with a great hand off to the next chat to continue the work." I run that anytime I approach 50% of the context window, as all models get dumb at that point. Then /clear, then /effort max just to be safe, then "please ingest chats/2026-01-01-00-00-what-we-did.md and proceed." It's a very purposeful custom /compress that works far better in my experience. If I ever hit auto-compress, I have failed as a Claude jockey.