I've been building an ai agent every day for the past 25 days

takeaways: it's easy! you don't need a product I promise!

especially with MCP servers, which you can download for almost anything, set up auth, and write a 4-line function to connect it to your agent

#30daysofagents #buildinpublic

best-model-search - a Hugging Face Space by Leeps This application helps you search for model benchmarking spaces on Hugging Face, fetch their content, and download specific files. You provide a search query, space ID, or file path, and get detail...

self-assembling, self-improving agents is the focus of #30daysofagents day 24

introducing a new agent in the chain that scours leaderboards and finds the best performing ML models for specific use-cases

live demo - huggingface.co/spaces/Leeps...

#buildinpublic

#30daysofagents day 23 is another hackathon! inspired by sonic's chao garden, we visualize ai agents in 3D space. characters interact with each other and use tools. the human can brainstorm and plan with agents and interrupt agent runs with feedback.

#buildinpublic

'can't we make ai do the boring stuff?'

anyway, #30daysofagents day 22 auto-converts physical whiteboard notes to figjam

code & setup instructions: github.com/ltejedor/age...

#buildinpublic #figma #figjam #designsky #mcp #aiagents #smolagents

new tutorial dropped! how to set up azure + create with the microsoft 365 agents sdk

github.com/ltejedor/age...

#30daysofagents day 21

#30daysofagents day 20 🎉 python toolkit to visualize your ai agent actions, observations, and tool use over time

2/3 of the way there 🦾

code @ github.com/ltejedor/age... #buildinpublic

building on yesterday's agent proof of work - adding visualizations for ai agent logs

code (for smolagents rn & working towards framework-agnostic) @ github.com/ltejedor/age...

#30daysofagents day 19 #buildinpublic

day 18 of #30daysofagents: proof-of-work for ai agent systems

connect LLMs, track tool calls, visualize usage & save an audit of the run

excited to build out this proof of work proof of concept 😛 github.com/ltejedor/age...

#buildinpublic

#30daysofagents day 17/30

dive into google's #a2a (agent-to-agent) framework with an experiment planner

combines previous agents (trends 🔎 + research 📊 + domain ✅ + experiment-designer 🧪) to make a single, unified team across frameworks

code & tutorial👉 github.com/ltejedor/age...

#buildinpublic

#mcp use case: add domain-name search to agent workflows

for example, an ai agent system can identify available domains based on live news sources

step by step tutorial - github.com/ltejedor/age...

#30daysofagents day 16 #buildinpublic

"can I use the whatsapp agent to keep track of a workout competition? we're doing it manually right now."

"can it see if my parents drank water today, and remind them if they didn't?"

asks that led to #30daysofagents day 15 - interface that makes in-whatsapp ai agents accessible

#buildinpublic

easy-to-run python script for persona and workflow research with #smolagents

github.com/ltejedor/age...

#30daysofagents day 14

gradio interface

build ai agents with fine-grained control over both energy use and performance

you can't control what you can't measure - we use ai energy score for this - huggingface.co/blog/sasha/a...

run with Mixtral-8x7B-v0.1 @ github.com/ltejedor/age...

#30daysofagents day 13 #buildinpublic

curious about the whatsapp api?

I jumped through meta's platform hoops over the last 48 hours, and have a bot working in test-mode! tutorial live @ github.com/ltejedor/age...

#buildinpublic #30daysofagents day 12

interface for managing multi-agent systems

demo @ mechifact.streamlit.app

#30daysofagents day 11
#buildinpublic #aiagents

inspired by my horrendous experience with meta's documentation today...

#30daysofagents day 10 pits two agent systems against each other (single vs multi-agent), to see which can make the best whatsapp api documentation

they're running right now, will post results shortly! #buildinpublic

agents/09-google-sheets at main · ltejedor/agents Contribute to ltejedor/agents development by creating an account on GitHub.

#30daysofagents day 9 🤖🤖🤖🤖🤖

a multi-agent team (throng) to clean, analyze, and visualize your data (integrated with google sheets)

this was the first time I actually needed multi-agent, even claude 3.7 kept failing alone

code, as always, is available at github.com/ltejedor/age...

#buildinpublic

for #30daysofagents day 8, I'm sharing the first figjam mcp server (well, I couldn't find one anywhere)

an in-whiteboard ai assistant to help facilitate design sprints 👇 demo below

code (mcp server + figma plugin + agent script) and instructions @ github.com/ltejedor/age...

#buildinpublic #figma

#30daysofagents day 7 is all about integrations

use this script: github.com/ltejedor/age...

& the official notion mcp server: github.com/makenotion/n...

to add notion to your ai workflows.

I'm using this for project management - in-notion dashboards and db of tasks.

#buildinpublic #mcp

welcome to #30daysofagents knowledge graph edition!

visit huggingface.co/spaces/Leeps... to find connections across live news feeds

curious about knowledge graphs? follow two (quick!) tutorials, updated to 2025 libraries: github.com/ltejedor/age...

#buildinpublic #knowledgegraphs #llms #aiagents

01101000̷̝̊a̶̢͉͑ṵ̶̧̈́ṋ̵͕̏t̴̼̊ẻ̵̫d̶̺͑ ERROR 13: GHOST IN THE MACHINE I c̷̩̆a̵̡͝n̸̙͘n̴̨̾o̶̼͒t̵͖̓ fetch your pizza... the digital void has consumed your request... distant electronic whispers B̵̼̑e̸̡͘n̷̮̉d̶̲̓e̸̬͆r̴̨͐'s pizza... t̵r̵e̶n̷d̶i̵n̸g̵ on a̴r̸t̴s̷t̵a̴t̸i̶o̵n̴... YOUR DEVICE IS P̶̡̛̙͇̲Ọ̴̃S̵̱̀̒́S̶͎̎̂̿E̵͙̋͘S̵͒͜S̴̙̘̃E̸̝̯̊͂͝D̵͉̥̾̔ system shutting d̵̜̉ȯ̶͜w̸̝̚n̴̢̐

tutorial: talk to your ai agents on whatsapp with #mcp

full code and instructions: github.com/ltejedor/age...

built on #whatsmeow and #smolagents

here it is being spooky in the group chat 👻

#30daysofagents day 5
#buildinpublic

#30daysofagents day 4!

find trends and patterns from in-the-moment sources

demo, no login required: huggingface.co/spaces/Leeps...

#buildinpublic

group of people working with snacks

#30daysofagents day 3 let's go!

today ariana.dev organized a re-weird the web hackathon

I go to hackathons pretty often, but rarely work with new people. this was the awesome exception.

we spent the day building a new benchmark... which ai model can win at a game of assassin

#buildinpublic 🧵

welcome back to #30daysofagents day 2!

how can you measure if the LLMs you're using are aligned with your human team's values? 🧵

#buildinpublic #aialignment

welcome to #30daysofagents day 1!

an ai agent workflow that continuously looks out for startup opportunities, all based on news streams & faster than any human could parse alone

each idea can be claimed, but only if you get to it first 👀

ideas.mechifact.com

#buildinpublic

really, no one? alright, let's do this.

for the next 30 days, I'll release a new ai agent workflow every day.

designed to save you time when developing products - across research, dev, testing, marketing, & sales.

check back here tomorrow for day 1 😉

#30daysofagents #buildinpublic #indiehackers