Stop wasting time and tokens by using Clean.
Joined March 2026
- Tweets 25
- Following 7
- Followers 308
- Likes 38
4 Photos and videos
clean retweeted
Jun 2
clean mcp slashes the tokens your coding agent burns finding code. now open source.
right now your agent greps "login," gets 200 matches, opens 15 files just to find 1 function. clean kills that whole loop with local, semantic search using lancedb for claude code / cursor.
repo: github.com/cleanmcp/clean-mc…
3
1
5
1,409
May 31
Hey u have to survive so clean can be big
May 31
did my first hike ever today. 6 hours in Sedona. nobody told me you’re supposed to EAT first
me: water a Monster
the mountain: absolutely not
And i burned 3617 calories yes u read that right my whoop went crazy
ended up in a CVS googling what an electrolyte is 💀
anyway. survived. SF next 🫡
263
May 28
So hop on now
May 28
Claude Opus 4.8 is now available in Clean.
It's able to handle long running tasks better than Opus 4.7 & still surpasses GPT 5.5 in frontend tasks.
1
140
May 28
You can use opus 4.8 with clean now
May 28
Tried Opus 4.8 Honestly its just hype not actual improvements just another “better benchmark” release and they degraded opus 4.7 so i cant use any good model
I will just switch to kimi at this point
101
May 28
I believe its a crazyy feature
May 28
our AI draws system architecture.
it also has zero patience for bad design.
we showed it a microservices diagram from a $500/hour firm. it redrew it in 90 seconds. fewer services. fewer failures. fewer lies.
the firm asked if we were hiring. we said 'no, but our AI is.'
🎨 @cleanailabs
210
clean retweeted
May 27
introducing: Clean Architect.
you drop a GitHub URL.
in seconds, you get a full interactive architecture diagram of that repo.
not a summary. not a README. the actual system - components, data flows, dependencies - laid out visually.
then you use that canvas inside your coding agent chat.
5
4
53
5,963
clean retweeted
May 24
introducing reference repos.
index any repo in clean. @ mention it whenever you need the context.
no cloud. everything runs locally.
we even tell you exactly how much storage and LLM usage each repo costs before you commit.
stop copy pasting your codebase into chat. just @ it.
4
4
18
16,777
May 21
We’re only improving each day. One shot an implementation.
May 21
we just killed @OpenAI deep research.
same prompt. same paper (Titans, Google). chatgpt got the equations wrong.
Clean is a local-first deep research tool, 7 parallel sub-agents, runs on your machine, reads your files.
full side-by-side ↓
1
1
216
May 19
we do work wonders.
May 19
no way, i just one shotted LoRA with clean. i was just testing out the deep research feature i was making, and it worked so good @cleanailabs
1
2
278
May 18
Nice
1
2
241
May 13
Clean works with your Claude Code and Codex subscriptions. Improve your productivity workflow with us.
1
4
221
May 13
They say you work better in clean environments!
3
142
Apr 30
We have finally launched our version 2: Clean Agent.
Clean Agent does:
- Automated Testing
- Smart Prompting
- Context Management
- Deep Research
End-to-end code generation with smart context management just for you.
Go to tryclean.ai
2
1
9
834