DEV Community: Mahesh

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.

How to Create an AI Agent using Google ADK | Python SDK

Mahesh — Fri, 04 Sep 2026 07:57:58 +0000

Google ADK makes it easy for you to create an AI agent. And it supports different language SDK to make your onboarding easier. Typescript, Golang, Kotlin, java and Python. In past I covered typescript and Golang.

Today I want to explain the Python SDK based agent creation experience I had with Google ADK. I created a video for you to check out. It includes an example of Agent chat interface for you to try out.

Python makes it very easy for you to create an agent. You can also glue the agent to variety of APIs and the libraries available in the Python ecosystem.

I decided to create an agent that makes it easy for you to check the answers from the Wolfram Alpha API. And in the fallback I made it easier for the LLM to make the query and get the answer. So the entire agent keeps functioning.

You can also find the use case of your choice. Create an UI or command line utility for the agent. And that would help you build the rest of the workflow around the AI agent.

Connecting with external APIs where you want to use the accuracy, answers and the variety of the fixing of the areas that would kind of help you build agent better as well. In that context I would say Google ADK along with any agentic coding tool can build your agents lot easier.

How to Create Text to Speech with NeuTTS and Hermes Agent

Mahesh — Tue, 01 Sep 2026 15:44:43 +0000

Hermes agent pretty much rules the Text to Speech world with it's approach of supporting them out of the box. Some of them are listed in it's list of providers. NeuTTS is one of them.

NeuTTS makes it very easy for you to turn your script into the speech audio. And a lot of sample native English and other languages supported too.

I created a video that would show you how to setup the NeuTTS and convert text-to-speech in Hermes Agent.

Hermes agent has the tools commands which would show you Text to speech libraries included. And you would find the NeuTTS listed over there. You just have to choose it, there are no API keys to add.

Another alternative is to choose the NeuTTS through github URL and ask the hermes agent to use it either through python package or clone the repo and learn from the sample.

It would take some time because VENV makes it difficult for the env setup and the packages to be run that quickly. And considering how the Hermes agent runs test before it can show it is done. You have to give it some time.

You can watch my video as I explained this setup to example process. Also have some sample text file ready as that would make TTS example quicker.

That's it. You have it setup and you are free to make the TTS experiments.

How to Connect Hermes Agent to mem9 AI Memory

Mahesh — Sun, 30 Aug 2026 08:48:48 +0000

Hermes agent is pretty much open towards the memory providers and it has plenty of them listed already. I also recently covered Hermes agent and Mem0 AI Memory post.

Now I recently integrated my Hermes agent with Mem9, which is another service which supports Openclaw and Hermes agent. It's free within limits, open source and have cloud instance too.

You can check out my video for managing the Mem9 AI memory and integrating it with the Hermes Agent.

How to get started?

First go to official website of mem9. You have two options either download it as an instance or login with gmail and github and get into the dashboard.

If you choose cloud route you would have to create an API key. And that API is what you are going to share with the Hermes agent. This would be one off operation and you'd get done. Save .env file with that API key.

You can also ask the Hermes agent to pick the API key and read from mem9 docs to set the Mem9 as the memory provider.

You would require to check the command hermes tools. From there you would be able to access the memory providers. Once you set that up make sure to also command hermes agent on what to take it to the memory.

For example,

Save my frontend design choices, personal life (safe events) are stored in the memory for you to access later. Do not store it locally. Store it in attached cloud provider for AI memory.

Make sure to also get rid of memories which are not safe to share. Because at the end you don't the providers of agents and memory use that against you as a customer.

That being said, mem9 is worth exploring and saving your memories.

How to Create AI Voice Agents using ElevanLabs API

Mahesh — Fri, 28 Aug 2026 15:57:15 +0000

When it comes to voice AI agents, there are many TTS providers out there. But if you are trying to build real time speech solution and if you wish to build a service bot or AI agent with voice capabilities it's expensive.

However there are some of the providers like ElevenLabs which offers some credits to build the apps on their ecosystem.

I created a video explaining how to Create Voice AI agent using ElevenLabs API.

Here's how I start for any new voice AI agent.

Lets say you want to build voice agent for your store. First come up with data on your store. So that agent can pick the inventory, script and the data related to store.

Use langchain or Google ADK. And this framework would build a basic agent. You can use the Vercel Eve framework to build a chat agent around this product. And you can use LLM which gives decent processing for the text to voice conversion on the go.

This is the frame of your solution. You can extend this into the prompt and build the app. App is how your get agents to work. Building agent doesn't mean just writing the code on Google collab and be at that stage.

Build the implementation with an app. That's how you learn to build Voice AI agent.

How to Connect Hermes Agent to Supermemory AI Memory

Mahesh — Wed, 26 Aug 2026 06:02:42 +0000

Hermes agent is pretty decent tool it can get the a lot of user tasks done. While there are some cases where you wish to have a memory and set of instructions. So that you don't repeat your prompts. You don't repeat your everyday automations.

This is where the memory comes in. By default Hermes agent has user.md and the memory.md. These two files hold that data. But you have limited control over that. You can extend this using external memory providers like Supermemory.

I have made a process video of how to setup Supermemory with Hermes agent. Give this one a look if you quickly want to set memory with Hermes Agent.

Supermemory gives you 5$ worth of the AI memory credits. Which would be more than enough for most of the Hermes agent usage.

You can use it to store your session data, ai agent's preferences, your stack preferences, your database and other memories even including personal.

This way you can filter out memories in such way that your workflow would be easier to put up with the local memory. And you can sync your memory with the cloud as well. This is something we can work with Hermes agent pretty easily.

Hermes agent is pretty good with the memory management and using external provider gives you more freedom to keep your rules and the memories fixed and let the agent learn from it saving you time on prompts.

How to Create an AI agent using Google ADK | Golang SDK

Mahesh — Sat, 22 Aug 2026 16:04:14 +0000

Google has created a framework for building an AI agent. This framework is called as a Google Agent development Kit. And it also has variety of the language SDK for the framework.

I recently covered the Typescript SDK and now I also covered the Golang SDK with the Google ADK. I made a video for you to check out the whole process.

In this particular Agent I decided to turn the framework output into the CLI agent. It made my life easier when I used this SDK and decided to create command line output to build the final agent.

In such agent development it helps if you sign up to a service where you can get the free AI inference API. Like say Kilocode or Openrouter or similar other type of the service. I also think you can use Opencode with their free models in the API.

In this tutorial I decided to use the Weather API and that API I managed to get access to certain APIs endpoint and some of the city data for the weather related output.

At the end I made this agent such that it answers most of the weather answers, and for it makes use of the LLM output where the weather API data ends. So both those options kind of contribute for the development.

I'd recommend you to check out the Google ADK and use any SDK of your choice - typescript, python and golang SDK. I feel Python and golang just are too good and quick to build the command line apps.