CEO at Graftcode, Creator of Javonet, Co-founder/CTO at SdNcenter, Software Engineer, Software Architect
San Francisco, CA
- Tweets 499
- Following 1,465
- Followers 220
- Likes 857
Przemysław Ładyński retweeted
this is one of those things that makes you realize how far dev tools have come.
❌ before: hours of AI integration glue code
✅ now: expose the methods already living in your codebase
@Graftcode turns ordinary public class methods into MCP tools through its gateway.
slick.
With Graftcode, you can expose your methods to LLMs without writing any additional layer.
MCP works out of the box, with support for JavaScript, C#, Java, and Python.
Less code. Faster integrations. More time building.
Link in comments 👇
Paid partnership (ad)
2
16
28
5,534
🚀 graftcode CLI just got even better!
Our CLI now supports downloading plugins!
Windows
iwr grft.dev/get | iex
Linux / macOS
curl -fsSL grft.dev/get | sh
Easily add Service Bus or RabbitMQ plugins, reference them in your Graft & Graftcode Gateway config.
Route all communication through queues, scale out your solution, while keeping remote service calls as simple local functions! 🔥
By removing integration-specific code from your codebase, you can adopt AI-assisted dev workflows in distributed systems — dropping token consumption by 50% and simplifying diffs in AI-created pull requests, as presented at WeAreDevelopers 2026.
graftcode.com
#Graftcode #DevTools #Microservices #WeAreDevelopers
2
7
115
Guys! What’s the best way to gather early adopters to evaluate new dev tool that literally just killed REST, gRPC, Thrift, SignalR, DAPR and GraalVM in one shot?
1
47
🚨 AI Coding has a HUGE problem… and nobody’s talking about it! 🔥
Microsoft gave thousands of engineers the most powerful tool (Claude Code) — and quietly canceled the licenses a few months later.
Uber burned through its entire yearly AI budget by April.
Even Nvidia (the ones making the AI chips) revealed their team’s compute costs now exceed what they pay employees!
Why? Because AI works too well. It generates code at lightning speed, but context, tokens, and integrations eat millions. Context windows get bloated, and you’re paying every time the AI has to re-learn your entire stack.
In this video, RoadsideCoder breaks it down with a real SaaS example (Next.js + Supabase + AI APIs):
• Why model selection is everything (not always Opus!)
• How context reconstruction eats 40-60% of your tokens
• Tools like Graftcode that actually help
If you’re building with AI agents — you NEED to watch this. Link in comments / below.
This isn’t FUD. This is 2026 reality. AI speeds up development 10x… but the bills do too. 💸
Who’s already feeling these “hidden costs” in their projects? Drop a comment 👇
#AICoding #AIagents #HiddenCosts #DeveloperLife #TechTwitter
1
2
3
50
Przemysław Ładyński retweeted
That’s why Graftcode is interesting for AI-assisted development.
The hard part is not only generating code.
It’s keeping the service context intact while frontend, backend, and cloud services talk to each other.
Graftcode is built for that integration layer: less manual wiring, fewer tokens burned, and backend logic available directly where you need it.
Beta quick-start:
academy.graftcode.com/quick-…
3
8
414
What you think guys about new approach to expose APIs I posted yesterday?
Make .NET service with just public methods callable remotely in 1s...
With @graftcode you can make any pure module remotely callable via strongly-typed client from any technology — in 2 commands:
1) Create empty project
2) Build it
3) Navigate to build output folder
4) Run command to get Graftcode Gateway
5) Run: gg.exe yourmodule.dll
Works with .NET, Ruby, Python, Java, JavaScript, Kotlin, Groovy, PHP, C# and more!
Every public method becomes remotely callable for web apps, mobile apps, microservices — and even exposed as MCP tool for AI!
To consume remotely:
→ pick your Package Manager
→ copy install command for your graft
→ run it on calling side and enjoy always up-to-date strongly-typed client!
🚀🚀🚀🚀🚀
Link in comments 👇
2
1
33
Make .NET service with just public methods callable remotely in 1s...
With @graftcode you can make any pure module remotely callable via strongly-typed client from any technology — in 2 commands:
1) Create empty project
2) Build it
3) Navigate to build output folder
4) Run command to get Graftcode Gateway
5) Run: gg.exe yourmodule.dll
Works with .NET, Ruby, Python, Java, JavaScript, Kotlin, Groovy, PHP, C# and more!
Every public method becomes remotely callable for web apps, mobile apps, microservices — and even exposed as MCP tool for AI!
To consume remotely:
→ pick your Package Manager
→ copy install command for your graft
→ run it on calling side and enjoy always up-to-date strongly-typed client!
🚀🚀🚀🚀🚀
Link in comments 👇
1
2
5
77
If you can do it your AI too! Imagine how simpler your system design would be, how much less work you would need and how much less tokens you would consume! 🚀
Try yourself now: academy.graftcode.com
2
3
22
Przemysław Ładyński retweeted
Your AI assistant generates a microservice in 15 minutes
Then you spend 3 hours wiring it into your stack and burn 50k tokens managing service communication
AI scales code not integration
@Graftcode fixes this and its free
43
216
219
127,562
Przemysław Ładyński retweeted
I deleted 30 lines of MCP server code. My Claude Desktop agent still works, thanks to @Graftcode🙌
Full story shared via @Medium : medium.com/p/i-deleted-30-li…
Tools I used in this exercise - @claudeai @Docker Desktop @antigravity
#mcp #agents #ai #graftcode #graftcodegateway
1
3
3
103
🚨 This Python dev deleted his entire MCP server code…
…and Claude STILL works perfectly.
He wrote ONE plain Python file with pure public methods.
No @mcp.tool() decorators.
No FastMCP imports.
No wrapper files.
ZERO extra lines of code.
That single file now powers:
✅ Full MCP tools for Claude Desktop
✅ Backend for Web & Mobile apps
✅ Public REST API
Same code. Zero boilerplate.
Fewer lines = way fewer AI tokens to generate it.
This is the biggest productivity hack for AI agents I’ve seen this year.
Full demo in the comments 👇
Would you delete your MCP boilerplate too?
#MCP #ModelContextProtocol #ClaudeAI #AIagents #Python #ZeroBoilerplate #GraftCode #Docker #LLM
1
2
2
106
Przemysław Ładyński retweeted
1/ I just deleted 30 lines of MCP server code.
My Claude Desktop agent still works.
Better, actually. 🧵
1
2
2
169
Suryansh Tiwari
Who tried this approach? Would you agree it could reduce tokens and simplify PRs in ai-assisted dev workflows?
teddit.net/r/betatests/s/TdY…
11
27
527
1,105,458