Set Up Hermes Agent for Free
Install a self-hosted AI agent and have it build a working web app for you.
Introduction
30 Second Summary
Imagine telling a computer exactly what app you want and watching it build the whole thing for you. No coding tutorials, no syntax errors, no Stack Overflow rabbit holes. Just one sentence typed in a terminal or texted from your phone, and a working app minutes later.
In this project, you will install a self-hosted Hermes AI agent on your machine, connect it to a free language model through OpenRouter, and give it a single instruction to autonomously build a working Pomodoro timer app.
What You'll Build
A fully configured AI agent that autonomously writes and runs code on your machine, producing a working Pomodoro timer app without you typing a single line of code.
By the end of this project, you'll have:
- A self-hosted Hermes Agent connected to a free AI Model.
- A working Pomodoro timer app built entirely by the agent from one prompt.
- A screenshot and conversation transcript proving an AI wrote working software for you.
- Secret Mission: Text your agent an app idea from your phone.
Quiz Yourself
Your score won't affect your project. This is just a gauge of your prior knowledge about the topic.
Before We Start
Before We Start...
Every tool in this project runs on your own machine. Hermes Agent is an open-source AI agent that can write code, run terminal commands, and build complete applications on your behalf.
Install Hermes Agent
In this step, you'll install Hermes Agent using its official one-line installer. The installer automatically sets up all dependencies, so you won't need to install anything else separately.
In this step, get ready to:
- Run the Hermes Agent one-line installer.
- Reload your shell and confirm the hermes command is available.
- Complete the Setup Wizard to connect OpenRouter and configure your tools.
Open your terminal and run the installer
The installer is a single terminal command that downloads and sets up everything Hermes Agent needs to run.
- Check if Git is already installed by running this command:
git --version
✔️ I see a version number
You're all set. Git is installed and ready for the Hermes installer to use. Continue with the install command below.
ⓧ Command not found
Git is not installed yet. The Hermes installer needs Git to clone its repository, so install it now.
macOS
- Run this command to install Git via Xcode Command Line Tools:
xcode-select --install
- A popup will appear asking you to install the tools. Click Install and wait for it to finish (this can take a few minutes).
Windows
Your WSL2 Ubuntu terminal includes apt for installing packages.
- Install Git by running:
sudo apt update && sudo apt install git -y
Linux
- Install Git by running:
sudo apt update && sudo apt install git -y
- Once the install finishes, confirm Git is working by running:
git --version
You should see a version number like 2.39.0. Any version is fine for this project.
macOS
- Press Cmd+Space to open Spotlight, type Terminal, and press Enter.
- Paste this command into your terminal and press Enter:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
Windows
Hermes Agent requires WSL2 (Windows Subsystem for Linux) to run. WSL2 gives you a full Linux terminal inside Windows, which is what Hermes needs to execute commands. If you already have it, continue below. If not, the tip at the bottom of this tab walks you through installing it.
- Press the Windows key, type the name of your Linux distribution (e.g. Ubuntu), and press Enter.
- Paste this command into your WSL2 terminal and press Enter:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
Don't have WSL2 yet?
Install it now with these steps:
- Press the Windows key, type PowerShell, then right-click it and select Run as Administrator.
- In the PowerShell window, run:
- wsl --install
- Wait for the download to finish. When prompted, restart your computer.
- After restarting, Windows will automatically open an Ubuntu terminal and ask you to create a username and password. Choose anything you like.
- Once you see a $ prompt, WSL2 is ready. Continue with the install command above.
Linux
- Press Ctrl+Alt+T to open your terminal.
- Paste this command into your terminal and press Enter:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
What does this command do?
This downloads and runs the official Hermes Agent install script. The installer handles everything automatically:
- Clones the Hermes Agent repository to ~/.hermes/hermes-agent.
- Sets up a managed virtual environment with Python 3.11, Node.js, uv, ripgrep, and ffmpeg.
- Adds the hermes command to your shell PATH.
The installer takes 1-2 minutes to complete. Once it finishes, the Setup Wizard launches automatically.
Walk through the Setup Wizard
Right after the install finishes, Hermes automatically launches the Setup Wizard. This wizard configures your provider, model, terminal backend, messaging platforms, and tools in one go.
If the wizard did not start automatically, launch it by running:
hermes setup
Choose Full Setup
The first screen asks how you want to set up Hermes. You will see three options: Quick Setup (Nous Portal), Full Setup, and Blank Slate.
- Type 2 or hit arrow down and press Enter to select Full Setup.
Do NOT press Enter on the default
The highlighted default is Quick Setup (Nous Portal). If you press Enter without typing 2, you will be routed through the Nous Portal path, which does not use the free OpenRouter key this project relies on.
If you accidentally selected Quick Setup, press Ctrl+C to cancel, then re-run hermes setup and choose option 2 this time.
Why Full Setup instead of Quick Setup?
Quick Setup connects through the Nous Portal, which bundles a paid Tool Gateway for web search, image generation, and browser automation. This project uses a free OpenRouter key instead, so Full Setup lets you wire up your own provider without any subscription.
Oops Headsup! Before configuring the wizard, you need an OpenRouter API key. Choose the tab that matches your situation.
✔️ I already have an OpenRouter account
- Navigate to https://openrouter.ai/keys.
- Click Create New Key.
- Give your key any name (e.g. hermes) and click Create.
- Copy the key that appears. It starts with sk-or-****. Keep this tab open or paste the key in a notepad.
ⓧ I need to create an OpenRouter account
OpenRouter is a unified API that gives you access to dozens of free models with a single key. No credit card required.
- Open your browser and go to https://openrouter.ai.
- Click Get API Key or Sign up in the top right corner.
- Create a free account using your email or GitHub login. No credit card is needed.
- Once signed in, navigate to https://openrouter.ai/keys.
- Click Create New Key.
- Give your key any name (e.g. hermes) and click Create.
- Copy the key that appears. It starts with sk-or-****. Keep this tab open or paste the key in a notepad.
Select OpenRouter as your provider
- Go back to your terminal. The wizard now shows a list of providers (OpenRouter, Anthropic, OpenAI, etc.). Select OpenRouter.
- When prompted for your API key, paste the sk-or- key you copied from OpenRouter a moment ago and press Enter.
Why OpenRouter instead of calling a model directly?
OpenRouter acts as a single gateway to 25+ free models. If one model goes down or hits a rate limit, you can switch to another without changing anything except the model name. One key covers them all.
Pick a free model
The wizard displays a list of available models. The free models are at the bottom of the list, marked with free labels.
- Press / to open the filter, then type free to narrow the list to free models only.
- Select any model whose ID ends in :free. Good choices include free variants of NVIDIA Nemotron, DeepSeek, Qwen, or Llama models.
- Press Enter to confirm your selection.
Which free model should I pick?
The available free models change daily, so pick whatever the list shows you. As an example, an NVIDIA Nemotron :free variant is a strong choice if it appears. Any model with :free in its ID will work without spending credits.
Hermes requires at least 64,000 tokens of context. If a model has less than that, the wizard will reject it automatically.
Choose the terminal backend
- The wizard asks where Hermes should run commands. The default is Keep current (local). Press Enter to accept it.
Why local?
Local means the agent runs commands directly on your machine. This is required so it can create and open the app file you will build later in this project.
Because it runs locally, Hermes will pause and ask for your approval before executing each command. This is normal. You will approve commands when the agent asks.
Skip messaging platforms
- The wizard shows a list of messaging platforms (Telegram, Discord, Slack, etc.) with checkboxes. Leave ALL boxes unchecked and press Enter.
- Do not toggle anything here. You will not need messaging for this project.
Configure tools for CLI
The wizard shows a checklist of tools to enable. Some of these tools require the paid Nous Portal gateway, so you need to uncheck them to keep this project free.
Press Space to uncheck these three tools:
- Web Search & Scraping
- Browser Automation
- Image Generation
Keep checked:
- File Operations,
- Terminal & Processes, and
- Code Execution.
- Press Enter to confirm.
Why uncheck those three tools?
Web Search & Scraping, Browser Automation, and Image Generation all require the paid Nous Portal gateway to function. Leaving them enabled forces extra provider-setup screens for each one and can cause a free model to stall trying to use a tool that is not connected.
Unchecking them skips those detours entirely and gives your agent a clean, focused set of tools that work out of the box.
Left those tools checked by mistake?
If you left Web Search, Browser Automation, or Image Generation checked, the wizard will ask you to choose a provider for each one. In every case, select the option labeled Skip or pick a free option. Never select anything marked [paid] or [subscription].
Handle the text-to-speech screen
- The wizard asks you to choose a text-to-speech provider. This screen appears regardless of your tool choices.
- Select Skip or accept the free default (Microsoft Edge TTS). Do not select anything paid.
- Press Enter to finish the wizard.
Installation complete. woohoo!
Confirm your setup with hermes status
The wizard is complete. Now reload your shell so the hermes command picks up all the new configuration.
- Reload your shell by running this command:
macOS
source ~/.zshrc
Windows
Your WSL2 terminal already picked up the PATH change. No action needed.
Linux
source ~/.zshrc
- Check what version of Hermes Agent you installed by running:
hermes --version
You should see a version number like 0.17.0 printed in your terminal.
hermes status
Your output should show OpenRouter as the configured provider and a model ending in :free with no authentication errors.
Seeing errors in hermes status?
If the provider is wrong or missing, re-run hermes setup and walk through the wizard again. Make sure your API key starts with sk-or- (not a plain sk-, which is an OpenAI key).
You can also run hermes doctor for automated diagnostics that check your config, keys, and dependencies.
help me fix my hermes status errors after running the Setup Wizard
If you're ever stuck, ask the NextWork community. Students like you are already asking questions about this project.
Hermes Agent is installed, connected to OpenRouter, and running a free model. Next, you will verify the connection holds up and learn how to switch models if needed.
Connect a Free Model via OpenRouter
Verify and Switch Models
You just connected OpenRouter and picked a free model during the Setup Wizard. Now it is time to confirm that connection is solid and learn how to swap models if one is not working well.
Free models on OpenRouter rotate availability and have rate limits (50 requests per day, 20 per minute). In this step, you will verify your setup, understand what those limits mean, and see how to switch to a different free model if yours goes down.
In this step, get ready to:
- Verify your OpenRouter connection with hermes status.
- Understand free-tier rate limits.
- Learn how to switch models with hermes model.
Verify your connection
You already ran hermes status at the end of Step 1. Run it one more time now to confirm nothing has drifted.
- Confirm your provider and model are still active by running:
hermes status
- Confirm the output shows OpenRouter as the provider and your chosen :free model with no authentication errors.
Want to switch models?
✔️ I'm happy with my current model
No action needed. Your model is connected and ready to go.
What are the free-tier rate limits?
Without adding credits, OpenRouter allows 50 requests per day and 20 requests per minute. If you hit the limit, Hermes will return an error. Wait a few minutes or switch to a different :free model with hermes model.
For this project, 50 requests per day is more than enough to complete every step.
What should I see?
The hermes status output displays your configured provider (OpenRouter), your chosen model name, and the status of your authentication. If everything is green or shows no errors, you are ready to chat.
ⓧ I want to switch to a different free model
If your current model feels slow or is hitting rate limits, you can switch to a different free model at any time without re-running the full wizard.
- Open the model selector by running:
hermes model
What does hermes model do?
This command opens the same provider and model picker you saw during the Setup Wizard, but without re-running the full wizard. It keeps your existing terminal backend, tools, and messaging settings. You only change the model.
- Select OpenRouter (it will already be highlighted if it is your current provider).
- Press / and type free to filter for free models.
- Pick a different :free model and press Enter.
- Run hermes status again to confirm the new model is active.
What are the free-tier rate limits?
Without adding credits, OpenRouter allows 50 requests per day and 20 requests per minute. If you hit the limit, Hermes will return an error. Wait a few minutes or switch to a different :free model with hermes model.
For this project, 50 requests per day is more than enough to complete every step.
What should I see?
The hermes status output displays your configured provider (OpenRouter), your chosen model name, and the status of your authentication. If everything is green or shows no errors, you are ready to chat.
ⓧ I already have Claude Pro/Pro Max
If you already have a Claude Pro or Pro Max subscription, you can connect it to Hermes through an OAuth flow. No manual API key creation needed.
- Open the model selector by running:
hermes model
- Select Anthropic via API or Claude Code from the provider list.
- Hermes shows the Anthropic credential screen with three options. Select Reauthenticate (use new OAuth).
- On the next screen, select Claude Pro/Max.
- Your browser opens automatically to the Anthropic authorization page. Sign in with the same account that has your Pro or Pro Max subscription.
- Click Authorize to approve access.
- After authorizing, the browser displays an authentication code. Copy this code.
- Go back to your terminal and paste the code when prompted, then press Enter.
- Hermes immediately shows a model selection list. Pick a Claude model (e.g. claude-sonnet-4-6 for quality or claude-haiku-4-5 for speed) and press Enter.
- Run hermes status to confirm Anthropic is now your active provider with your chosen model.
How does billing work?
Your Pro or Pro Max subscription includes a monthly API credit allowance. Usage through Hermes counts against that allowance the same way the Claude chat interface does.
Status shows errors?
Re-run hermes model and double-check your API key. You can also run hermes doctor for automated diagnostics.
The free tier is limited to 50 requests per day and 20 requests per minute. If you have already hit the limit, wait a few minutes or select a different :free model.
help me diagnose why hermes status is showing errors after connecting OpenRouter
If you're ever stuck, ask the NextWork community. Students like you are already asking questions about this project.
Your agent now has a brain. Next up, you will have your first conversation with it and confirm it can reason about its available tools.
Start Your First Conversation
Your model is connected and Hermes Agent is ready to talk. Before you hand it a real task, you need to confirm the connection actually works.
In this step, you will launch Hermes in its terminal interface, send a quick test message, and verify the agent can respond and reason about its own capabilities.
In this step, get ready to:
- Launch Hermes Agent in TUI mode.
- Send a message and confirm the agent replies.
- Exit the session cleanly.
Launch Hermes in TUI mode
The TUI (Terminal User Interface) is the recommended way to interact with Hermes. It gives you a full chat interface right in your terminal with live tool activity, modal overlays, and mouse support.
- Launch Hermes in TUI mode by running this command on your terminal:
hermes --tui
Why --tui instead of hermes chat?
Hermes has three ways to start a conversation, all using the same agent and tools:
- hermes or hermes chat launches the classic CLI. It is a traditional text prompt, great for quick one-off queries or scripting.
- hermes --tui launches the TUI (Terminal User Interface). It gives you a full-screen interface with live tool-call activity, collapsible panels, mouse support, and non-blocking input.
We use --tui because you can watch the agent's tool calls stream in real time while it works. When the agent builds your app in the next step, you will see each file write and terminal command appear live instead of waiting for a wall of text at the end.
You should see a welcome banner appear with your model name, available tools, and loaded skills.
The banner groups information into collapsible sections. The Tools section is expanded by default, showing the tools the agent can use (terminal, file operations, etc.).
What is the TUI banner showing me?
The banner is your at-a-glance dashboard for the session. It confirms which model is active, what tools the agent has access to, and what skills it has loaded.
If the banner shows your chosen free model name and a list of tools, your setup is working correctly.
Banner not appearing or seeing an error?
If you get a connection error mentioning your API key, re-run hermes model and paste your key again. Make sure it starts with sk-or-.
If the model fails to load, try selecting a different :free model with 64K+ context. Free model availability changes daily.
Run hermes doctor in a separate terminal for automated diagnostics.
help me troubleshoot why Hermes TUI is not starting properly
Send a test message
Now that the TUI is running, you will send the agent a simple message to confirm it can receive prompts and respond coherently.
- Type the following message into the chat input and press Enter:
Say hello and tell me what tools you have available right now.
The agent should reply within a few seconds, listing its available toolsets such as terminal and file operations. This confirms both that your model connection is working and that the agent can reason about its own capabilities.
Why this specific message?
Asking the agent to list its tools is a quick way to confirm two things at once. First, the model is responding (your API key and provider are working). Second, the agent is aware of its tools and can reason about them.
In the next step, the agent will use these exact tools (terminal and file operations) to build you an entire web application.
Getting empty responses or errors?
If the agent gives empty or garbled replies, the free model may be overloaded. Exit with Ctrl+C, then re-run hermes model and pick a different :free model.
The free tier allows 50 requests per day and 20 requests per minute. If you have already been experimenting, you may have hit the limit. Wait a minute and try again.
help me figure out why the agent is not responding to my messages
Exit the session
- Exit the TUI by pressing Ctrl+C or typing /quit and pressing Enter.
You are back at your normal terminal prompt. The agent responded, your connection is confirmed, and you are ready for the real task.
Your agent is alive and talking. Next up, you will give it a single instruction and watch it autonomously build you a complete web application.
Have the Agent Build You an App
Your agent proved it can think and communicate. Now it's time to give it a real job.
You'll type a single prompt describing a web application. The agent will autonomously write all the code, save the file, and open the finished app in your browser.
In this step, get ready to:
- Create a clean working directory for the app.
- Give the agent a build prompt for a Pomodoro timer.
- Verify the finished app works in your browser.
Start a fresh session in a clean directory (new terminal)
The agent needs a dedicated folder to write files into. Creating a fresh directory keeps the app isolated from the rest of your system.
- To open a New Terminal, on macOS: Click New Tab with Profile – Basic from the menu (or press ⌘ + T) to open a clean session.
- On Windows press window key to open your search bar. Type Terminal and press Enter to open a new terminal.
- Create a project directory and launch Hermes inside it by running these commands:
mkdir ~/hermes-app && cd ~/hermes-app
hermes --tui
What do these commands do?
- mkdir ~/hermes-app creates a new folder called hermes-app in your home directory.
- cd ~/hermes-app moves your terminal into that folder so the agent writes files there.
- hermes --tui launches the agent's TUI (text user interface) chat session.
- Confirm you see the welcome banner showing your model name, available tools, and loaded skills.
Don't see the welcome banner?
- If you get hermes: command not found, run source ~/.bashrc (or source ~/.zshrc for zsh) and try again.
- If mkdir fails with a "file exists" error, the folder already exists from a previous attempt. Run cd ~/hermes-app and then hermes --tui directly.
Help me launch Hermes Agent in a new project directory.
If you're ever stuck, ask the NextWork community. Students like you are already asking questions about this project.
Give the agent your build prompt
A good prompt for an autonomous agent is specific about deliverables and constraints. You'll tell it exactly what to build, what rules to follow, and where to save the result.
- Paste this prompt into the Hermes chat (or adapt it for a different app idea like a tip splitter, streak tracker, or typing speed test):
Build me a Pomodoro timer app. Requirements:
- Single file: index.html with all CSS and JavaScript inline
- No external dependencies, no CDNs, no frameworks, no internet required
- Clean, modern UI with start/pause/reset buttons
- 25-minute work timer, 5-minute break timer
- Visual countdown and audio alert (use the Web Audio API for a beep)
- Save it to this directory, then open it in my browser
Why these constraints?
Each requirement solves a potential problem. A single file means no build step. Banning CDNs and frameworks means the app works offline with zero dependencies. Specifying the Web Audio API for the beep avoids the agent downloading a sound file.
The more precise your constraints, the less likely the agent is to take a wrong turn. Autonomous agents work best with clear boundaries.
The agent will begin making tool calls. You should see it using its terminal and file-writing tools to create index.html. This typically takes 2-5 minutes depending on your model.
What are the approval prompts?
As the agent works, it will pause and ask you to approve each command it runs. When you see a prompt like Allow this command? [y/n], type y and press Enter.
You will do this several times as the agent writes and saves files. This is normal and expected. It is a safety feature that ensures the agent cannot run anything you have not approved.
💡 Be patient with free models
Free models on OpenRouter can be slower than paid ones. Let the agent work through its tool calls without interrupting. You'll see progress messages as it writes code.
Verify the app in your browser
The agent will attempt to open the file in your default browser automatically. If it does not open, you can open it yourself.
- If the browser did not open automatically, open the file manually:
- To open a New Terminal, on macOS: Click New Tab with Profile – Basic from the menu (or press ⌘ + T) to open a clean session.
- On Windows press window key to open your search bar. Type Terminal and press Enter to open a new terminal.
- then run this command
macOS
- Open the file in your browser by running:
open ~/hermes-app/index.html
Windows
Since you are running Hermes inside WSL2, you can open the file from your Windows browser.
- Open File Explorer and type \\wsl$ in the address bar.
- Navigate to your Linux home folder, then open hermes-app/index.html by double-clicking it.
Linux
- Open the file in your browser by running:
xdg-open ~/hermes-app/index.html
- Confirm the app works: click the start button, watch the countdown begin, and verify the pause and reset buttons respond.
App not working or has external dependencies?
If the agent added CDN links or frameworks, tell it in the chat:
Remove all external links. Inline everything. No CDNs, no imports, no fetch calls.
If a specific button does not work, name the exact problem. For example: "The start button does not start the timer. Fix it. Remember: all CSS and JS must be inline, no external links."
If the agent loops or gives empty responses, the free model may be struggling. Switch models by exiting with Ctrl+C, running hermes model, selecting a different :free model with 64K+ context, then re-running hermes --tui.
Help me fix my agent-built app.
If you're ever stuck, ask the NextWork community. Students like you are already asking questions about this project.
You just had an AI agent build you a working application from a single prompt. Next up, you'll export the full conversation transcript to preserve a record of every tool call the agent made.
Capture Your Proof
Export Your Agent's Session
AI agents are not black boxes. Unlike a casual chatbot conversation that vanishes when you close the tab, agent sessions record every decision the model made, every file it wrote, and every command it ran. That record is inspectable and exportable.
In this step, you'll export your session as a JSONL file, then feed it back to the agent and ask it to generate a visual build report. You'll finish with a styled HTML page that documents exactly how your app was built — ready to open in a browser tab right next to the app itself.
In this step, get ready to:
- Export the full conversation transcript from your build session.
- Verify the exported file exists on disk.
- Have the agent generate a visual build report from the transcript.
Export your conversation transcript
Hermes stores every session locally on your machine. You just need to find yours and export it to a file you can keep.
- Return to your terminal (not the hermes --tui) or run a new terminal
- List your recent sessions by running this command:
hermes sessions list
You should see a table of sessions. The ID column shows a timestamp-based string like 20260630_193346_f20c7a.
- Find the most recent row in the table. This is your build session.
- Copy its ID value. You will paste this into the next command.
- Copy the session ID from the ID column for your most recent session and enter it here: your session ID from hermes sessions list
- Export the full session transcript by running this command:
hermes sessions export session.jsonl --session-id [[SESSION_ID="your session ID from hermes sessions list"]]
What does this command do?
hermes sessions export saves the full conversation to a JSONL file in your current directory. It captures your original prompt, every tool call the agent executed (file writes, terminal commands), and all the generated code.
For example, if your session ID is 20260630_193346_f20c7a and you choose session.jsonl as the filename, you would see: Exported 1 session to session.jsonl.
- Confirm you see a message like Exported 1 session to session.jsonl. The file saves to your current directory.
Don't see your session?
If hermes sessions list shows no sessions, you may have exited the TUI before it saved. Start a new session with hermes --tui, re-send your build prompt, and try again.
If the export command fails, double-check that you copied the session ID exactly as shown in the list (no extra spaces).
Help me export my Hermes Agent session transcript.
If you're ever stuck, ask the NextWork community. Students like you are already asking questions about this project.
Verify your exported transcript
The export command should have saved a file to your current directory. Let's confirm it's there.
- Confirm the exported file exists by running this command (use the same filename you chose in the export command):
ls -lh session.jsonl
You should see your file listed with a size of roughly a few dozen KB (for example, 39K). This JSONL preserves the full session record of your agent's work.
Why is this file valuable?
The JSONL file is a structured record of your entire agent session: your messages, the model's responses, every tool call, and token counts. It makes the agent's work fully transparent.
Professionally, this matters because you can audit exactly what the agent did (no guessing), debug issues by replaying its reasoning, and share the transcript with teammates as reproducible proof of how something was built.
Generate a visual build report
A raw JSONL file is proof, but it is not something you would show someone. Let's have the agent read its own transcript and turn it into a polished HTML page you can open in a browser.
- Launch a new Hermes session in the same directory of your pomodoro app by running this command:
hermes --tui
- Paste this prompt into the chat:
Read session.jsonl in this directory. It contains the full transcript of an AI agent session where you built me an app.
Generate a file name it build-report.html that visualizes the session as a styled build report. Requirements:
- Single file, all CSS inline, no external dependencies
- Dark theme with clean modern styling
- Sections: my original prompt, a timeline of every tool call the agent made (file writes, terminal commands), and the final result
- A footer that says "Built autonomously with Hermes Agent"
- Save it to this directory, then open it in my browser
What is happening here?
You are using the agent's own file-reading tool to ingest the transcript, then asking it to generate a new artifact from that data. This is the same pattern used professionally to auto-generate changelogs, sprint reports, and documentation from structured logs.
The agent will read the JSONL, parse the tool calls, and write a styled HTML file. This typically takes 1-3 minutes. Approve any tool-use prompts as they appear.
- Once the agent finishes, confirm the file exists by running:
ls -lh build-report.html
- Open the build report in your browser by running:
macOS
open build-report.html
Windows
Since you are running Hermes inside WSL2, open the file from your Windows browser.
- Open File Explorer and type \\wsl$ in the address bar.
- Navigate to your Linux home folder, then open hermes-app/build-report.html by double-clicking it.
Linux
xdg-open build-report.html
You should see a styled dark-themed page with your original prompt at the top, a timeline of every tool call the agent executed, and a "Built autonomously with Hermes Agent" footer at the bottom.
Report looks plain or is missing sections?
If the report is unstyled or missing the timeline, tell the agent in the same chat: "The build report is missing the timeline of tool calls. Read session.jsonl again and regenerate build-report.html with all sections."
If the agent cannot read the file, make sure you are running hermes --tui from the same directory where session.jsonl lives. Run pwd to confirm.
Help me get my agent to generate a build report from my session transcript.
If you're ever stuck, ask the NextWork community. Students like you are already asking questions about this project.
Secret mission
Text Your Agent from Telegram
You have an AI agent that builds apps on command. But you have to be at your terminal to use it. What if you could text it an app idea from your phone and come back to a finished project?
Clean Up Your Resources
Clean Up Your Resources
Decide whether to keep your resources running, pause them to come back later, or delete them entirely. This project runs entirely locally with a free API tier, so there are no ongoing costs.
Resources you used:
- Hermes Agent installation at ~/.hermes/ (free, local files only).
- Project files at ~/hermes-app/ (HTML apps built by the agent).
- Telegram bot (Secret Mission, free).
What about your OpenRouter account?
Your OpenRouter account is on the free tier with no credit card attached. It has no idle cost and no expiration. You can leave it as-is for future use.
Keep everything running
No action needed. Choose this if you want to keep using Hermes Agent for future projects.
- Hermes stays installed and ready. Run hermes --tui any time to start a new session.
- The OpenRouter free tier has no idle cost or expiration.
- Your Telegram bot stays active. Message it any time after running hermes gateway run.
- Your HTML apps in ~/hermes-app/ remain on disk and can be opened in any browser.
Pause - I'll come back to this later
Nothing is actively running after you closed the TUI and stopped the gateway. There are no background processes to pause.
- If you left hermes gateway run running in a terminal, press Ctrl+C to stop it.
- Everything else is already idle. No processes, no costs, no expiring resources.
Delete - I don't want to use this again
Remove all project resources and start fresh. Follow each step below.
- Remove your Telegram bot by opening Telegram, messaging @BotFather, sending /deletebot, and selecting your bot from the list.
- Delete the project files by running:
rm -rf ~/hermes-app
- Uninstall Hermes Agent and all its configuration by running:
hermes uninstall
What does hermes uninstall leave behind?
The hermes uninstall command removes the Hermes binary and code, but preserves your configuration and session data in ~/.hermes/ in case you reinstall later. To remove everything completely, also run:
rm -rf ~/.hermes
- Confirm removal by running:
ls ~/.hermes
You should see a "No such file or directory" message, confirming everything is gone.
That's a Wrap!
That's a Wrap!
Nice work! You just had an AI agent build you a working web application from a single prompt. You wrote zero lines of code yourself.
You've learned how to:
- Install and configure Hermes Agent, a self-hosted open-source AI agent that uses terminal and file tools to write software autonomously.
- Connect a free LLM via OpenRouter with no credit card and no ongoing cost.
- Direct an agent to autonomously build a complete web application from a single prompt, and capture proof with a screenshot and exported session transcript.
- (Secret Mission) Connect a Telegram bot to your agent and trigger a remote app build from your phone.
Does it say 'Still tasks to complete!' at the bottom of the screen?
This means you still have screenshots left to upload, or questions left to answer!
- Press Ctrl+F (Windows) or Command+F (Mac) on your keyboard.
- Search for the text Return to later.
- Jump straight to your incomplete tasks!
Still stuck? Ask the NextWork community. Students like you are already asking questions about this project.