we've shipped taskmaster v0.23 🚀 → official npm i task-master-ai Visual Studio Code extension → manage your taskmaster workflow visually → introducing scope-up & scope-down tools → dozens of QOL improvements & fixes → we're hiring 🐹 follow + retweet + bookmark 👀👇

1 introducing the official taskmaster VSCode extension get it here: → visual task management → multi-tag and context support → ai-powered task updates → config management → quick AI actions we'll be heavily investing into this moving forward!

it's an experimental release with a few rough edges if you have any trouble getting it to run, consider: → go to user settings → type `taskmaster mcp args` → remove --package=task-master-ai → make sure you `npm i -g task-master-ai` nothing like testing on prod amirite

2 introducing scope-up & scope-down → strength levels to choose adjustment intensity → --prompt flag to specify complexity adjustments → available in CLI, MCP & @code extension! a much requested feature for even more control over your tasks without over/under-building

3 many QoL fixes and improvements made with love → fix for tasks not found when using string IDs → tag-specific complexity report detection → and more! as usual, we rely on your feedback to release features you care about and improve features to further streamline things

will be posting some news this week to explore the future of taskmaster → what comes next? → what are we focused on right now and why → how to bring the rest of your team into the workflow → we're hiring 4x senior engineers right now dms are open to talented & hungry devs

that's it! v0.23 changelog: → npm i -g task-master-ai@latest → MCP & VSCode extension auto-updates → join → more til next time, friends vibe on!

help taskmaster reach more people by retweeting the first tweet and sharing it with your friends who are tired of getting stuck coding with ai make sure to follow if you bookmark!

@TaskmasterAI @code Woo! Thanks for letting me contribute! Appreciate you & the team for bringing it to life!

@TaskmasterAI @code well done david! 🤝

@TaskmasterAI @code This a great work of art 🫡 Code rabbit in cursor presently have a feature where the extension sends instructions about the comments to the llm chat interface. That might be something to do look into, you mentioned something like this briefly in the video

@TaskmasterAI @code that’s right! there is a way

@TaskmasterAI @code I look forward to the update. Thank you for your amazing work! 🫡

@TaskmasterAI @code Any future plans to focus on token efficiency? Over time this will become critical to manage.

@TaskmasterAI @code token efficiency is quite good already - anything specific you’re referring to?

@TaskmasterAI @code Specifically thinking about ways to track efficiency and ways to stay within subscription limits. I.e. CC Aug28 changes. Beyond what ‘ccusage’ provides.

@TaskmasterAI @code got it - awareness of the current context window usage esp if the client provides it would be a great idea

@TaskmasterAI @code the most passionate and active MCP server development I've seen, great efforts!

@TaskmasterAI @code You really doing amazing work !!! Nothing but gratitude for your contribution to my day to day building 🙏🏿

@TaskmasterAI @code vibe on david!

@TaskmasterAI @code We want to integrate taskmaster directly into @usekonsole Can we?

@TaskmasterAI @code @usekonsole Yep will work out of the box

@TaskmasterAI @code Can you tell me how is Taskmaster different or better than Roo code with tasks list?

@TaskmasterAI @code Sure - roo code produces its tasks on the go based on your prompt which implies surrendering control to the LLM and having it decide how things will be implemented this isn’t far off from just vibe coding one prompt at a time

@TaskmasterAI @code But I can manually edit or converse on revising what’s in the tasks list?

@TaskmasterAI @code you can - which would pollute the context window and eventually lead to context drift (the llm forgetting parts of what you said or hallucinating) its much more scalable to commit your context window to a single task at a time and give the llm tools to read/write from context

@TaskmasterAI @code Can you tell me how is Taskmaster different or better than BMAD method?

@TaskmasterAI @code I don’t know much about it From what I gather, there’s a difference in that Taskmaster doesn’t create a bunch of architecture documents Everything lives in the tasks and subtasks

@TaskmasterAI @code Amazing 🔥🔥🔥 I expect with the scope up and scope down cmds I will end up creating a smaller number of larger tasks. or is there a way to do it across multiple small tasks?

@TaskmasterAI @code yep you can scope-up/down multiple tasks at once (comma separated)

@TaskmasterAI @code nice, I will be using that a ton

@TaskmasterAI @code Great job! I appreciate how you deliver exactly what users want. Keep it up!

@TaskmasterAI @code we will!

@TaskmasterAI @code Is it only for vscode or also for cursor/windsurf/trae? Is there a way to use taskmaster with warp agent?

@TaskmasterAI @code All of them :) MCP on Warp works great

@TaskmasterAI @code this is the biggest "smaller launch" I've seen in awhile looks really hood. I was having difficulty reviewing and scoping things up and down, was finding it difficult to keep taskmaster on the right track if I needed to tweak the generated tasks. this looks so intuitive.

@TaskmasterAI @code Getting the extension. Retweeting

Do you have a guide for using it with Claude code? Not just installation, but considering they have their own task list system, and now sub agents. I love taskmaster, but I’m wondering if it’s getting redundant, and I should just lean all the way into Claude code. Need a Delta of value prop.

@TaskmasterAI @code i’ll put something together

@TaskmasterAI @code Yessss great point here. The drift is real.

@TaskmasterAI @code Supply chains arent fragile, they're rigid...adaptability is the missing link, esp. now w/ shifting consumer trends.

@TaskmasterAI @code @eyal you hiring? claude code engineer, Miami, FL.

@TaskmasterAI @code You need some kind of benchmark, does taskmaster really outperform claudecode or w/e

@TaskmasterAI @code hey @EyalToledano Couple of things: 1.Can we use OpenRouter key to access models, i don't want to add a perplexity and an anthropic key AND anthropic rate limits are so annoying 2.i told taskmaster a description of my project in CMD and cursor went off and built it..right away

@TaskmasterAI @code 1 yes, openrouter is fully supported 2 vibe on!!!

@TaskmasterAI @code Not sure if some advanced config is needed for OR but api key in mcp's + .env doesnt work error in next message

@TaskmasterAI @code run task-master models —setup pick an openrouter model right now your model is set to claude (by default)

@TaskmasterAI @code thanks! :) turns out i had an old version installed somewhere deep in npx !😐