31/03/2026 5:10 pm
Topic starter
Notifications
Clear all
Agent Skills Forum
3
Posts
4
Users
0
Reactions
58
Views
0
Im finally getting to start this new project for a local client where I have to build an agent that can basically automate a whole research workflow for their trading firm. I only have about 3 weeks to get a prototype running so im really hyped but kinda stressed about picking the right stack. Most of what I read says Python is the go-to for anything AI but then I saw some guys on Reddit arguing that TypeScript is way better for complex state management and building out the actual tools the agent uses. My logic was that Python has all the libraries but I dont want to hit a wall if the agent needs to be super snappy. Does it really matter if I stick to Python or is TS actually the secret sauce for advanced skill dev?
3 Answers
11
31/03/2026 5:45 pm
Just caught this thread and honestly, you're hitting on a real debate in the agent space right now. I've been using TypeScript for my last few agentic workflows and I'm really satisfied with how it handles complex tool definitions. If your trading firm client needs something that feels snappy and reacts fast to streaming data, TS is definitely not a bad shout. Python is great for the data science side, but Node's event loop is basically a dream for handling a dozen API calls at once without blocking the whole process. Here is why I'm leaning TS for your 3-week sprint:
- Type safety is a huge lifesaver when you're passing weird JSON objects between different agent steps. You catch way more bugs at compile time instead of waiting for a 2-minute agent run to crash.
- If you go this route, definitely look at LangGraph.js State Management Library v0.0.15 because it gives you that fine-grained control over the agent loop that's harder to get in base Python scripts.
- For the backend, running Node.js v20.11.0 LTS Runtime with Fastify Web Framework v4.26.1 is incredibly efficient for getting a prototype live quickly. I've got no complaints about library support either since most major providers like Anthropic Claude 3.5 Sonnet API LLM have killer JS SDKs that are arguably better than their Python counterparts. You'll probably find that the real secret sauce is just how much faster you can iterate when the types are doing the heavy lifting for you. It keeps the state management from becoming a total mess as the agent skills get more complex tho.
10
31/03/2026 5:55 pm
Re: Python for agent dev... I agree. One might want to consider:
- Using Pinecone Vector Database Serverless to scale costs.
- Be careful with context windows to avoid expensive overages.
2
31/03/2026 5:10 pm
Honestly, I've been using Python for agent development for years and I'm quite satisfied with the results. It works well and I've got no complaints regarding speed for research tasks. For your timeline, I recommend using LangChain Python Framework v0.2.0 alongside OpenAI API GPT-4o Model. These tools make building complex agent skills very straightforward. Python is basically the best choice to hit that three week deadline.
