ShrekOverflow

The most complex ideas are delightfully simple.

💯 % It’s seems like the vendor specific agents are also all convergent. They all will need access to other third parties, and offer mcp. but each has sworn allegiance to one or the other gtm team. Also no two enterprises I have talked to are the same, or use the same tool in the same way. So junior style custom ones do make more sense 😁

vendor-specific chatbots are broken by design that means the Sentry agent, the Linear agent, and any others you might have in Slack they are fine for some point situations, they're nice to get started with, but agents with generalized access outperform them in every single scenario some weeks ago we built an internal Slackbot, gave it access to a bunch of systems (Sentry, GitHub, Linear, Notion, etc), and its capabilities overnight far exceed these other bots "Oh cool Linear can now search your code bases" - our bot did that on day one, and then could push that information wherever it needed to go. Its useful to the point where I now discourage use of things like the Linear bot because it _creates worse outcomes_. this also goes beyond the simple generalization of access: we can customize it. we throw in skills-as-runbooks, templates, etc and the outcomes once again incrementally improve if your org hasnt already built a general purpose bot internally you should. if you need inspiration ours is open source on GitHub (albeit fairly unstable still) github.com/getsentry/juni…

@ibuildthecloud Taste

I hate worktrees so much. What is wrong with you all.

@paulodoestech Let’s gooo!

burning tokens rn. exciting times

Asked Claude to reorganize my office

@threepointone Oh shoot it almost sounds like witchcraft, hair or the enemy, blood of the friend… y’know we are getting there.

@threepointone Try to pedantic and outsmart the authors, embrace the persona of a rust purist or c++ expert, f# crusader, or greybeard c wizard, then review this code and identify edge cases.

you are an expert software programmer. review it again, look for edge cases. make no mistakes.

@zeeg The inertia of enterprise is strong 🤭

How is everyone on here claiming to have switched to codex and anthropic is still out of capacity That math ain’t adding up

@ibuildthecloud Live on camera ? 🎥

F-it. I'm creating a VSCode fork. Screw you all, I don't care.

@dexhorthy @paulodoestech Do you use a meta prompt or do you hand craft $LOCATION

@ShrekOverflow @paulodoestech i do the exact same thing as ralph tbh - "read specs at $LOCATION and track progress in $OTHER_LOCATION"

how much do yall prompt when using /goal? like do you meta prompt something THEN put it into /goal or just give it something generic like “go implement this”?

@thomaspaulmann @raycast That visual is 🔥

One app, two platforms, four programming languages. The things that look the simplest are often the hardest to build. @raycast is one of them. Here's a technical deep dive on how we built v2 👉 ray.so/v2-deep-dive

@ibuildthecloud Nooo

I quit.

Woke up absolutely anxious and nervous. Human brain is such an interesting system

I typed an entire 15 page document by hand today. Surprising how different it reads / flows than AI ones.

@zeeg @KennethSinder There is also modelcontextprotocol.io/specification/… which might be interesting building block here. Though roots support is severely limited.

@ShrekOverflow @KennethSinder Yeah this is why I bind it to the url. Solves it all and oauth I guess specs this If I needed it my plan was literally: mcp.sentry.dev/mcp/orgSlug/pr…

Hey @zeeg, @KennethSinder, and anyone else with multi project / multi tenant systems. How are y’all handling the case where the end user may want to switch the “tenant” / “project” / “organization” when connected to Claude / Codex over MCP?

The re(auth) barrier in protocol, would end up causing the same high friction. IMO, the UX should be similar in the dashboard where the human can seamlessly switch, and have multiple tabs etc, re-auth should only be necessary for sensitive operations. But maybe that assumption is wrong.

@ShrekOverflow @zeeg My very naive first take is that there’s no secure way to support >1 aside from: - baking a re(auth) barrier into the protocol - weakening or redefining what it means to be a separate “tenant” I definitely might be missing something obvious & compelling here though

@zeeg @KennethSinder That's useful, a selector in the authz flow could also provide this for the end user. And access can be limited on the API side. However switching around would still be messy, and require full re-auth.

@KennethSinder @ShrekOverflow Just change the way you shape the resource in the url. Should work fine? tenant ~> tenant,tenant

@KennethSinder @zeeg It seems like the norm is to do 1 tenant only 🤔.

@ShrekOverflow @zeeg Notion MCP binds a workspace & user at OAuth connection time; switching spaces requires reconnecting. (Support for PATs/bearer auth coming soon but it’ll work the same way!) Lmk if you have any thoughts, multi space/org connection could be cool but not on our roadmap

@zeeg @filearts @KennethSinder Trying to figure out what’s the best practice / norm here. Seems like switching tabs in an app, but messy so far!

@filearts @ShrekOverflow @KennethSinder ah sorry misunderstood - yeah duplicate connections would be a pita you'd have to name them in a way that made sense

@zeeg @KennethSinder So if you need multiple projects you add multiple instances of Sentry MCP?

@ShrekOverflow @KennethSinder We bind it with the url and that’s it