DEV Community: Mahesh

How to Add Zep AI memory to AI Agent

Mahesh — Thu, 15 Oct 2026 15:35:01 +0000

In this series of adding AI memory to the AI Agent. I decided to find more AI memories. And the goal was to focus on the services which have the cloud API service.

Take an example of the AI memory service which would make it easier for you to add the memory through cloud API and sync the memories for you to access later on the dashboard.

I found one such service called Zep AI memory. This allows you to setup the AI memory with the AI Agent. I decided to show you the entire process of how I did it.

Few things you should know before you do the same like me. Make sure that you have a use case that is under limit so much so that you don't end up triggering the free tier limits.

Most of the serious use case would benefit from the Zep memory in your AI agent with pro tier. So use that directly instead of doing the window shopping. Something you would find yourself with the difficult financially then free tier is good enough.

I mostly used ADK and Langchain for these video series. You are not limited for that. You can ask your AI coding agent to use mastra, crew AI and other ai agents and it won't have much issues either.

Do make sure to have a local sqlite database file in case if your usage go above the cloud API of the Zep service with the free plan. In such case the local file storage would do in the downtime scenarios.

How to Web Search with Tavily API & Hermes Agent

Mahesh — Tue, 13 Oct 2026 17:08:02 +0000

I have been working with Hermes agent a lot. Each week I try to do something with Hermes agent. And I make sure to document it. As with more usage of AI the more you would start to forget.

So recently I got my hands on the Tavily API. They offer free credits for the web search and the data extraction. And so I decided to use their API with the Hermes agent.

Below is the video that I created to document how you can use the Tavily API, do some web search, get summary and work with Hermes agent on processing such data.

Tavily API is similar to other search and extract services out there. They have dedicated support for agents. So you can also use this API very easily with cursor, and other IDEs.

You can also use their MCP and set them up. But the command line usage with the hermes agent would require you to setup the API part in .env file or some other way you have to create the menu option in the CLI.

I have basically used this to do the stock research. As the dividend stocks are my thing. I make sure to use them for some of the strong researches on the dividend portfolio. And I make sure to invest into them for long term.

My work with Hermes agent was to document this process and present to you. In short to show you how it all works and what you can do better than me. And how it would help you to build your features around the AI Agent.

How to Add Hindsight AI memory to AI Agent

Mahesh — Thu, 08 Oct 2026 15:56:01 +0000

AI Memory is going to be a new database for the coders. Not that it is meant to connect with the product or app. The role of that database is going to be the connection between the agent that works and the coder.

Hindsight is one such offering from the Vectorize. You would find the AI memory in both local setup and the cloud service. Which would be easy to work and setup with the AI agents.

I recently tested the Hindsight AI memory and it's worth exploring too. You can check out my process video here. It's something to check out.

You have two options - setup hindsight local or use Cloud API. Say you choose the hindsight local, that would make the data local to you and your control on that data.

So if you are worried about data leaving the computer, then use the local option for the Hindsight. And that would be a good direction for the local. In such case it makes sense for the growth in general how to work with such AI memory and build your apps.

Here in this example. I created an AI agent and allowed the hindsight to store the data. And I showed the step by step usage of the AI agent adding memory, accessing it, removing some context etc. It works smooth.

If you are building any AI agent then definitely take a look at the Hindsight AI memory and worth exploring with the AI agents.

How to Web Search with Firecrawl & Hermes Agent

Mahesh — Tue, 06 Oct 2026 13:41:58 +0000

Web is all about data. And the data you have ability to scrape and make sense of. Take example of Amazon data for products. If you are managing shop yourself you can learn from that data for pricing and inventory.

For scraping data there are plenty of libraries and online APIs. Take example of Firecrawl which makes searching the web, extracting data and exporting data easier.

I recently created a video where I used hermes agent with the Firecrawl API to build a process to search the web and extract data.

Firecrawl also gives you enough free credits for you to test out the scraping or web search workflow. Hermes agent can create python script for you where you can use the API.

And then give it a process, so that it can create a script to check the web, do the process of tasks, then extract data.

You can even do this for google search keywords, ads, youtube titles and endless tasks that Firecrawl can do for you. It is something you would easily do this with idea of your own.

Once your quota exceeds the genuine usage, you can go ahead and get the paid upgrade. I'd say most of the time you would be fine in the free quota itself. But as your hermes agent use case increase things would take speed.

You can check out Firecrawl and see if that it can do the automation workflow for you through the browser based automation.

How to Add Cognee Memory to AI Agent

Mahesh — Thu, 01 Oct 2026 16:59:43 +0000

AI agents are growing in terms of everyday use cases. However there is always an issue of consistency. More tasks they have to iterate the more you feel like supervising them.

This is where AI memory and the guardrails and policies come into picture. But how would you get them to remember the successful output? This is where memory comes in.

I recently learned about Cognee, which is one of the platforms that you can use for the AI memory and use them in your AI agents.

I have done a process video of how to use the Cognee AI memory and how to implement the memory in the AI agent. This would be something you may want to explore too.

Why should I use memory? Simple. When you want the agent to remain in context. It needs to access past use cases. It wants to know what are your preferences and the facts associated with your product workflow.

By doing this you get to understand where to use the memory in the agents. When to use them and when to not use them. This seems to be the only way for the agents to build context consistently.

That being said, try out cognee memory with the AI agents next time and see where they can be effective with your use case.

How to Automate Browser with Browseruse & Hermes Agent

Mahesh — Tue, 29 Sep 2026 16:23:38 +0000

Hermes agent makes most of the API based servcies glue into itself very quickly. And you get to work through them just with the prompt. I know on dev.to a lot of people are too much fundamentalist that they prefer code over no-code approach.

However I have learned how to automate better with Hermes than writing from scratch myself. So I found an API which is browseruse which allows you to do browser automation, take screenshot, automate browser for data extraction.

I created a process video for Hermes agent to be integrated with the Browseruse API.

You get 5$ worth of credit. I know lot less than the firecrawl and the browserbase. So that would be one of the blocker. But if you have means to pay for their API you can go ahead and try them out with initial credit.

And if you happen to like their service you can go ahead and build your own service around the same if you can. But paying is worth it if you are into automation. What I have found is that firecrawl and browserbase are enough for most of the tasks.

In case of the browser use the best use case is the pdf export and the screenshot. If you only want these two things then this can be a good API to explore. But apart from that I would say API would do most of the tasks there.

Have you explored any browser automation APIs with hermes agent? What's your experience so far?

How to Create Search AI Agent with Langchain & Tavily API

Mahesh — Thu, 24 Sep 2026 15:10:16 +0000

Langchain is one of the best AI Agent development frameworks. Between Google ADK and the Langchain you can find the use case wise popular framework being Langchain.

I decided to explore Langchain with some of the use cases like building search AI agents as one of the starting point. Langchain if used with any coding tool, makes it easy for you to build agents.

Take example of the video above where I have made a process on how to create an agent which would use the Tavily Search API. Which we can use to scrape web for data and build insights.

So building insight based bot is lot easier too. You can start with the sources where you wish to search. You can also let the AI agent figure out what data to give to the web. And then it would explore and find the right insight for you.

So how did I created this search AI agent?

Use langchain as an agent framework.
Use Openrouter for LLM models
Use Search AI Agent like Tavily and use API
Create CLI utility for search agent use case.

There are different types of the search agents. Like say stock search, finance insight of the companies, company insights for P&L and many use cases like this.

Tavily is a good search AI that would help you adapt to your use case. And it can even help with 1000 or so credits with which you can use the prototype for development.

Create a prompt based on what type of the search AI agent. For example you can explain how your CLI based agent you wish to search for and what type of insight you want.

You can then ask the agent to export the search insight into CSV and the excel spreadsheet. This way you would find the agent becomes more useful too.

Check the video above to see my process to build this Search AI Agent.

How to Automate Browsers with Hermes Agent & BrowserBase API

Mahesh — Tue, 22 Sep 2026 15:30:38 +0000

Hermes agent has a section called tools. This is where the "Web & Search" and also the "Browser" option is given. Where you get to use the external tools and APIs specific to the browser automation is being offered.

I recently did the connection of Hermes Agent with the Browserbase. You can check out my setup with the Browserbase and the Hermes agent for automating the browser.

BrowserBase a cloud service for data extraction, which makes it easy for you to automate browser using AI API. This makes it easy for you to do the data extraction, automation testing, screenshots and the PDF export.

In short here's how you set it up.

Method 1: Open Hermes agent tools section. Add the browserbase API by accessing the "Web and Search" options and then ask normally to do your automation, data extraction and other tasks.

Method 2: Another option is that Hermes Agent can read the .env file. Where the API key can be stored. And let the Hermes agent watch that particular file and then use it for the automation.

I am using the method 2 for the API which are not supported by the Hermes agent officially. And this would be possible for the development of small scripts to the apps. It makes sense to use the API the right way depending on how you plan code.

Browserbase is pretty handy and they have decent free limit for the API too. So if you are trying to use it for browser automation and the scraping it's worth exploring.

How to Create Search AI Agent using Langchain & Exa API

Mahesh — Thu, 17 Sep 2026 15:07:47 +0000

If you are developing an AI agent you would come across variety of use cases that would make it easier for you to save time, automate tasks and replace manual efforts.

Take example of the Search AI agents, which can take the data and scrape content from the web. It also can help take screenshots of the web pages. And then export the content to pdf and other types of the data formats.

I created Search AI Agent using Exa API and Langchain. I have made detailed process video to show you how you can do the same. This is use case specific content, and not code jam or code explanation session.

Exa.ai is an AI search service which would allow you to data search, agent API and variety of other API services they are offering through their platform.

Langchain was chosen because it has support for the variety of the AI inference providers. And also other features are included in the framework.

I decided to make use of the Startup news and Startup financials to be searched through this agent. And it is done in the command line. So you get the option to export such to csv or other data file format.

I'd say that if you are building command line utility then with langchain with saving the summary to the csv and other text format would be much better format.

If you are building agents, build UI and the use case around them, they may appear app but more you add iteration the stronger they become.

How to Automate Browsers using Browserless & Hermes Agent

Mahesh — Tue, 15 Sep 2026 14:28:58 +0000

Most of the automation on the web is through browser automation. Especially when it comes to scraping, data mining, and variety of the data extraction tasks.

I started using Hermes agent and run most of the tasks through it. It makes it easier than the n8n and other complicated workflows which can be painful to setup.

I recently started automating and perform certain browser automations through browserless. I also made the browserless connect Hermes Agent through API. Take a look at my process to set things up.

What does Browserless do better?

I had used firecrawl in the past. I may have to review it again. But browserless I used to export the data and the web page to the PDF and also to take the website screenshot.

You have certain amount of credits that you can use for the automation. Take example of the Hermes Agent with the tasks to extract certain type of the data and then make the output stored on the CSV. Which we can do with it easily.

Hermes agent can take your prompt then create a python script and then run any of your automation when you schedule it. And it can also store the output, gauge performance and it keeps on going with the learning from failure.

Considering the API is officially supported. I would recommend you to connect the Browserless API with the Hermes Agent and automate your small to large browser automation tasks.

How to Create AI Agent using Langchain & Gen AI SDK

Mahesh — Thu, 10 Sep 2026 15:32:46 +0000

Langchain has been making lives of developers easy with their agent framework. Making AI agent is now lot easier due to agentic coding and the framework from langchain. By adding additional extensible frameworks you even more those agents powerful.

Take example of Google's Gen AI SDK which can be plugged with the Google ADK and Langchain. I recently tried the Langchain noticed how easy it is to build agent.

A lot of coding part is already being taken care of agentic coding. I decided to build use case based AI agents. Here's my video on how to build such simple AI agent with Langchain.

How did I made it?

I started by asking the Gemini on creating a simple ai agent. My goal was to give it Openrouter Key and the models. And from there onwards my goal was super specific.

Goal was to build an desktop assistant that can be used as an agent convo within command line. I didn't went for something fancy with web based chat wrapper.

This agent can be simple to use and you can chat like any other AI model. And forget which model is behind. So I try to make it simple to learn more about agents.

If you build agents then go with use case based agents. This way you learn how to use the AI agents in real life.

How to Connect Hermes Agent to Hindsight AI Memory

Mahesh — Tue, 08 Sep 2026 16:43:02 +0000

AI memory has spawned out in the industry of AI agents. And it has managed to be useful for variety of use cases. A lot of use cases are specific to keeping user profiles, facts, preferences and other useful memorization tool cases.

I recently tested Supermemory and many other similar platforms where AI memory providers are given free tier. Vectorize too has released the Hindsight as hosted and cloud free tier. So I decided to try out the cloud version.

The video below basically shows the process of how to setup the Vectorize's hindsight AI memory. This makes it pretty decent to use for agentic purpose.

Hermes Agent has officially added the support for the Hindsight too. You can get the cloud API key and input into the memory setup for the Hermes agent.

You can also host the Hindsight locally. And that also gives you the access to the hindsight's memory capacity. I have used it to create certain interfaces for medical data storage. You can have your own features and the agent can do the storage with it.

Hindsight is a good AI memory storage. You should try both local and cloud API. Either way you are not going to be disappointed. Its effortless to setup for the Hermes agent.

For those who use Hermes agent would find Mem0 and Hindsight pretty easy to use and learn.