AI Agents For Trading (Free and Opensource)

All right, so I open sourced my AI agent

for trading here and all you have to do

is go to the link to get it. You can

just hit star and fork fork if you want

the code. Anyways, I wanted to go over

this the trading functionalities within

this package. There are a ton of agents,

20 different 20 plus different live

agents, trading agent, strategy agent,

risk agent, copy agent, whale agent,

sentiment agent, listing arb agent,

etc., etc. I'm a quanto. I focus mostly

on algo trading, but of course I will be

building all types of different agents

here. I think this is obviously a new

frontier. I mean, I don't have to tell

you that

we're going to keep building here and

then MCPs are here now and MCPs may lead

into something interesting with chat GPT

the app store. We'll see if that hits.

If not, I think agents are still the

way.

We're going to have AI doing stuff for

us. I think we can all agree on that.

So, trading, why wouldn't they? Let's go

ahead and work on this and go through

here. I want to go to the agent section.

It's kind of a little rundown of the

code. If you're on the read me, go read

through the read me. You can understand

what we've built. Then go to the source.

Everything will be in the source src.

You can play with all the models here.

You know, you have Claude, Deepseek,

Gemini, Grock,

O Lama, if you want to run local models,

OpenAI, of course, XAI, and this will

explain the whole model factory.

Everything is written out, so it's super

easy to run. Go to the SRC here. And I

wanted to go to agents today to talk

about those trading agents.

So, let's go ahead and learn about them

from scratch, dude. Let's learn about

them. Trading agents.md is going to be

where I'm reading from here now. Table

of contents here you can see. Anyways,

Mundav's trading system uses multi-agent

architecture where specialized AI agents

work independently or together to

analyze markets, generate signals,

manage risk, and execute trades. So,

agent categories, core trading, market

intelligence, token discovery. Okay, so

we have multiple agents and you saw

where to get those, right? It's in the

agents folder. We we're in it right now.

Trading strategy and risk, execute

trades and manage portfolio, market

intelligence, sentiment, whale funding,

liquidation chart, provide market

context, and then token discovery like

the sniper, Solana, and copybot.

Sniper and copybot. I wonder what this

does. Buy new opportunities. Of course,

none of this is financial advice. I'm

not financial advisor. Please take this

seriously. Pause this video right now.

read that whole thing or screenshot it

if you're on your phone.

So core trading agents here, these are

the mains agents that make trading

decisions and execute them. So the

trading agent, trading agent py, the LLM

decision maker. So this is pretty basic.

This is where we first got started. Just

a decision maker based off an LLM. So if

we go there, that should be called the

trading agent. The trading agent. Okay,

let's go ahead and put her to the side.

And you can see here we have a prompt.

Your Munda's AI trading assistant

analyze the provided market data and

strategy signals to make a trading

decision. So, I gave it a bunch of

information plus the data

and

you can see everything.

Let's close this for now.

You could see the prompts here. They're

all here. You can go through this code

on your own time.

What it does is pure AI analysis of

market data using claude or whatever. No

hard-coded rules. LLM decides

everything. Manages portfolio allocation

across all tokens. analyzes technical

indicators

and it can incorporate strategy signals

from other agents. How it works? Collect

market data. So it collects market data

and then it has the LLM analyze them.

Handles exits first always to manage

risk. LLM decides portfolio allocation

and then execute entries. So key methods

analyze market data. LM analysis

allocate portfolio LM decides capital

distribution

allocate allocations executes

handles exits configuration here I was

using haiku at the moment

AI model

max position percentage what I want to

do here is I want to say hey

at trading agent

Since we just implemented X AI,

their Grock model the four, I want you

to do the same thing with um this

trading agent.

Also make sure that the uh switchability

with the models is at the top of the

code with parameters. Thank you.

Okay, so that's the trading agent. It's

just like super simple like hey you

decide AI and when I tested it didn't do

too great. So uh there needs to be more

than that. So that led me to the

strategy agent. So, strategy agent, the

rulebased strategist

loads custom strategies from our

strategies custom folder. Each strategy

generates buy, sell, or nothing. AI

validates all signals before execution.

Can work independently or feed signals

to the trading agent. So, I had a

conversation with this guy here. His

name is Ernest Chan. You can go to the

podcast here. It should be live.

And he does something.

>> Please, please, please stop using chat

GPT like Google. Disappear for a month

and do this.

>> He does something.

>> Welcome to corrective AI.

>> Anyways, corrective AI is hey, give the

AI the strategy and then or yeah, give

the AI the signal from the strategy and

it gets to decide if it wants to run it

or not. So opposed to like predicting

price or something like that. Everybody

wants to predict price with AI. That may

not be the way. I can't tell you yes or

no. But

you can test that all yourself.

So here I wanted to figure out.

Let's keep going through here. So it

loads in some custom strategies. So

that's pretty cool because now we can

add custom strategies and then have AI

decide from those strategies. See what

I'm saying? And you see where this is

going. Okay, step one, we just say, "Hey

AI, just tell me what to do. Here's the

data. Let me know long, short, or

nothing." And that's elementary. So

let's go ahead and try a next step.

Okay. Well, here's a strategy. And now

it's triggered. Let me know if I should

take it or not.

It's interesting interesting ideas.

So then we have the risk agent, the

safety guardian. What it does, it's a

circuit breaker. Stops trading when

limits hit. Monitors max loss or gain

over configurable time frame. AI powered

override decisions for edge cases. So

it's like having a risk control plus AI

helping out with it. Minimum balance

protection runs first before running

anything else. So, it's looking at max

loss, max gain, and hey, maybe have an

AI say, "Oh, this is actually this is

running fast.

Let's not take profit." But then again,

that might get you smoked. So,

you can read through all of this here,

but um I wanted to see the first couple.

Dez the whale watcher monitors BTC open

interest changes detects whale activity

large OI moves AI analyzes if the moves

are significant announces. Let's get Dez

going. Let's get Dez back up in here.

Dude,

get Dez back. Okay. And you got we got

XAI here. Perfect. Let's see if it still

runs.

Beautiful. Beautiful. Beautiful. What is

that?

Oh, it's collecting data.

Bird eye API key.

Okay. Okay. This is working, which is

good.

Okay. I'm not going to let it keep

running. It's about to about to buy

something probably.

execute an entry.

Okay, so we just have to tweak that in

order to the EMVs to fix that. And then

let's go ahead and say get add dot get

commit slash um I can do this later

actually.

Updates

get push. Okay.

Let's scroll through here after I get

Dez going, the whale agent. Let's go

ahead and get her going. Where is she?

Well,

do you see it? I don't see it.

Whale agent.

agents.

Where'd she go? She's right there, dude.

Right in your face.

Sheesh.

This is using Deep Seek here. Copy path.

activate

activivate

Tflow and then Python run it.

Okay, it's running.

beautiful. Let's keep going through here

and seeing some other other agents. We

have funding agent. Okay, this one's

just telling me announces via voice when

thresholds hit.

This is interesting actually.

I might run this as well.

Funding agent.

Liquidation agent here. Luna the

liquidation monitor tracks liquidation

volume longs or shorts. Detects sudden

liquidation spikes. AI analy an analyzes

if liquidation signals reversals. So, I

started building these for voice because

if I get them built that way, then it's

super easy to be like, "Okay, we have

all the signals here. Put in a trade on

XYZ exchange." I want to look into that.

Matter of fact,

please look through this entire GitHub

and don't print any sensitive keys or

anything, but I want you to explain the

exchanges we have connected. So, I know

right now this connects

to at least Solana blockchain because I

just saw that in the trading agent, but

um yeah, look through the whole GitHub

and yeah, give me an update on that

chart analysis agent. What it does

generates candlestick charts with

indicators, uses AI vision to analyze

analyze chart patterns, supports

multiple time frames. So you see what

I'm doing here, right? I'm building out

the all the base agents. And then if you

want to try trading with this, you just

combine the chart analysis agent with

the trading agent. Trading agent knows

how to trade, right? So same with the

liquidations. Oh, you want to try some

liquidations? Okay, the liquidation

agent has all the data. It's just giving

it to me in voice right now. And then it

can easily be tested.

Chart analysis agent. Okay, cool, cool,

cool. Token discovery agents. These are

interesting. These were back in when

Solana was really going crazy.

Watches for new Solana tokens in real

time.

Solana agent. The Solana token analyzer

analyzes recent token launches using

mundage criteria.

MCAP

minimum liquidity.

So criteria is market cap. Okay.

Liquidity

token age.

Copy by agent analyzes current copy

portfolio positions.

Okay, this was cool actually.

Damn, there's so many ideas. There's so

many paths.

How agents work together. So the main

loop integration here,

here's how you can orchestrate all

agents together.

So you can bring in the risk agent, the

strategy agent, the trading agent, the

sentiment agent, the whale agents,

funding agent, liquidation agent. That's

a ton of different ideas

that could work together however you see

fit.

Initialize all agents here. Track

signals from intelligence agents.

So always check risk first. So got the

risk agent. Okay. Collect market

intelligence.

Strategy signals. Trading agent.

Information flow. Quick start guide.

Configure your system.

Tokens to trade. Position size. I may

have just made this for Solana, dude.

We'll see. Bird eye. Mundave.

run individual agents

trading agent strategy. Yeah, I think I

probably did because

yeah, that makes sense cuz I was going

to do this all on chain.

I don't know. We'll see. We'll see. I

just asked that question. Per agent

overrides best practices.

Risk plus trading agent rule-based

hybrid.

Always run risk first. Cost optimization

monitoring. No trades executing. Okay.

Yada yada. This is a lot of a lot of

slop you added on to this dude. A lot of

slop.

Okay. Sick dude. Remember trading

involves risk.

Please read the

this right here because that's super

duper important. So, I feel like that's

a pretty good overview of how the

trading agents work.

You just go to source here and you have

access to all of these agents here right

off the bat. I've got a lot I want to

update to this and I'm going to continue

to do that and I'll continue to make

videos as well just to show you step by

step along the way. So, you know, if

you're

interested in this GitHub here,

then make sure to, you know, subscribe

or follow or like or whatever you need

to do so you see the next one. But I

think that's about it for here.