Agent mode llms.txt .json
Display

hermes-xray

// Type a simple prompt, then watch the agent loop expose intent, tools, evidence, and verification.

hermes-xray is a standalone browser observability lab for a Hermes-style tool-using agent. Enter a prompt and inspect the model-visible path: prompt intake, context construction, loop policy, tool dispatch, verification, and persistence. It is independent from the Agent Runtime walkthrough and does not share that project's repository, portfolio slug, or Interact targets.

The lab shows operational evidence only. It does not expose hidden chain-of-thought, secrets, credentials, or private session fragments.

User prompt
01 Prompt intake normalize request + session
02 Context build memory, skills, history
03 Loop policy tools + safety rails
04 Tool dispatch execute and observe
05 Verify + persist check result, save state
Inspectable trace
Runtime:Static HTML, CSS, and JavaScript in the browser
Exports:llms.txt and hermes-xray.json
Interactive lab

Send a simple prompt. Inspect the runtime path.

Turn a plain request into an inspectable Hermes-style agent trace. Use the stage buttons to move from intake through verification without leaving the page.

Prompt: Review the portfolio-site deploy checklist

The trace shows operational evidence: intent, selected tools, file/code landmarks, and verification. It deliberately does not show private reasoning tokens.

Loop1 / 3
Prompt tokens8
Tool calls0
Stop reasonintake
01Prompt intakeNormalize request + session source
02Context buildMemory, skills, history, constraints
03Loop policyChoose tools + safety rails
04Tool dispatchExecute, observe, adapt
05Verify + persistCheck result, save useful state

01 Prompt intake

    Boundary: this view can show model-visible prompts, loop state, tool calls, token accounting, and observable rationale summaries. It does not expose hidden chain-of-thought or secrets.

    Source-shaped landmark

    
      
    
    
    

    
    
      
    
        
    Model-visible prompt stack
    
        
    
      
    
      
    
        
    Observable event stream
    
        
    
      
    
    
    
  
    

  
    
    
    Boundary
    
    
    What hermes-xray shows, and what it withholds
    
    
    
      
    
        
    Shown
    
        
    Exact user input, model-visible prompt and context summaries with redaction, loop stages and stop reasons, tool names/arguments/results, verification notes, and illustrative token or latency metrics.
    
      
    
      
    
        
    Withheld
    
        
    Hidden chain-of-thought, secrets and credentials, private memory or session fragments, fabricated live model output, and any claim that this browser lab is a production Hermes Agent deployment.
    
      
    
    
    
    
    
      Separate from Agent Runtime: Agent Runtime compares two open-source runtimes side by side. hermes-xray is a different project: an interactive observability console with its own GitHub repo, Vercel app, portfolio card, and demo route.