DEV Community: Mahesh

How to add AI memory in Hermes Agent | SQLite Memory

Mahesh — Tue, 03 Nov 2026 17:48:09 +0000

Hermes agent has support for multiple AI memory providers. I have covered a lot of services which are officially supported and also recently started exploring the ability of traditional SQL AND NoSQL databases to be a good memory system for Hermes agent.

You don't have to have a database with the vector search support. Instead a traditional SQL engine is more than enough for Hermes agent.

Why? Because by default there are 2 markdown files like user.md and memory.md files that hold the memory for the Hermes agent.

Take an example of SQLite database which I am currently using for the hermes agent. By default the database of the hermes agent is called holographic, which is SQLite database file. But if you want to create your own that is also supported.

SQLite makes data storage simple. And even better is how you store the memories from the Hermes Agent. Some of the time such storage of the files in the database file is simple.

And then making apps around your memories from this file is even more easier. You can even create the diagrams, charts etc out of such facts, preferences too. I think the use cases are endless.

I covered this in my video pointed above. You can check that out and setup the SQLite as a memory for Hermes agent.

How to Add Letta Memory in AI Agent

Mahesh — Sat, 31 Oct 2026 07:19:55 +0000

AI memory is one of the interesting things to practice when it comes to AI agent development. AI memories can teach you a lot about vector database, storing facts and preferences, and also managing agents for the agentic development.

I recently learned about variety of AI memory services like cognee, hindsight, etc. And also tried one new platform named Letta which has both SDK, CLI and also platform itself with inference provider.

I tried the SDK to build an AI agent with memory. Like connecting the Letta AI SDK with my AI agent and letting it store the AI memory data. It is free to start with your own Inference keys. I created a process workflow video for the AI agent development with Letta AI memory.

Letta memory you can test out using both desktop as well as the CLI interface. I have learned to use both in my examples later. But here I covered only the AI agent instance of it.

I also learned how to make the most out of the AI agent development with AI memory recently with variety of the business and the enterprise scenarios.

If you build apps and the interfaces for internal enterprise use then using Letta can be a good option to try out. Considering they have the SDK in open source this can be a good option to start with for your agentic memory.

How to Web search with Exa API and Hermes Agent

Mahesh — Tue, 27 Oct 2026 15:04:22 +0000

When it comes to the web search, there are direct APIs that we can use with the Google and other search APIs. But when it comes to the AI agents there is going to be dedicated search APIs that we can use.

Take example of Exa API, they are offering 1000 credits that can be used for the web search. And then you can use it with your own agents or you can use it with automation agents like Hermes Agent.

I have documented my process of integrating Hermes agent with the Exa API and manage to try out some examples. So if you happen to build something on similar lines you may want to check my below video if you want.

Exa is one of the popular web search agent. And you can use it for building variety of tools and the services. I have used it for the keyword research. Some of the other use cases are also the financial insight & research.

Hermes agent can be used for recurring searches and the web search where the decent amount of the development can be done with the APIs and this is where the Exa's free credits would benefit in your context.

Hermes agent is worth exploring. It has set of features worth trying out and web search APIs would make things easier for search. It's a good experience letting agent do the work for you and unless you have heavy usage 1000 credits are enough!

How to add mem0 Memory to AI Agent

Mahesh — Thu, 22 Oct 2026 15:55:35 +0000

AI agents these days need memory. They need memory for context in the project. They need memory so that they can do better processing and understand the small details of your AI Agents.

This is something I have noticed with almost every project where AI agents are meant to carry out some detailed work. And AI memory happens to build on such details by remembering their past iterations.

I recently tried mem0 AI memory platform. And decided to use it for the development. It's free tier is good enough to get started with. I created video that documents the process.

I tried this with both - Google ADK agent and the Langchain agents. I think with the langchain you get plenty of market providers in the AI memory niche. Overall I would say that use mem0 for starters and see how much painless it is for making use of it.

There are also other memory providers which I recently covered like Hindsight, Cognee and also other memory providers. But in terms of simplicity and free tier I feel the mem0 platform is pretty good.

You can also use these same memory platforms with the hermes agent. I have covered them separately in the past which you can check from my feed. Most of them have the free tier too.

Mem0 is free and also you can use it for your personal and the startup projects and setup and dashboard is pretty handy. Worth exploring if you are attempting to try out the AI agents with the AI memory.

How to Web Search with Hermes Agent & Parallel API

Mahesh — Tue, 20 Oct 2026 16:20:55 +0000

I have been covering the web search AI Agent APIs for some time now. I try to find APIs which make it easy for the developers to use it with AI based applications. And they can use it to extend their content.

Hermes agent is getting better. Far better than openclaw ever could get better. You can use Hermes in almost every place including the android phone. And VPS server.

So getting the web search is not that difficult. All you have to do is make sure to use the API from the parallel and ask the Hermes agent to add it as a plugin so you can switch for the web search.

I decided to use such APIs often for the stock research. Because you can add your strict criteria's and also make sure to let the Hermes agent with any models focus on those criteria's. Which kind of make things easier for your filtering of the results too.

Apart from that you can see how the Parallel and many other similar API services offer enough free credits which makes it easy for you to develop and work with the searches and you get decent data.

Also make a habit to export those results to CSV or any other document depending on your search results. Which in general kind of makes this enough for testing grounds. I feel most of such Search Web APIs are worth exploring even in the paid tiers.

If you happen to find any new Web search APIs for the Hermes agent, do let me know in the comments.

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.