Live Narrative Demo · Web-to-Agent Interface Compiler

Arachne

Arachne

Scout · Map · Compile · Govern

Watch Arachne transform a normal human website into an agent-operable surface: static pages, native APIs, observed JS traffic, extracted entities, and form actions become governed MCP resources and tools.

Run Demo See Outcome
AgentSiteModelNormalized map of pages, resources, forms, routes, confidence, and policies.
agent-manifest.jsonMachine-readable tools, resources, action schemas, and approval requirements.
llms.txt + MCPAgent-readable guidance plus structured resources and tools.
Nemesis-readyAdversarial surface map for policy bypass, prompt injection, and risky actions.
Current Live Platform

More than a crawler now.

The production backend supports readiness scoring, paid reports, preview compiles, domain ownership verification, hosted MCP delivery, drift checks, artifact history, and customer-ready bundles.

Compile modes

fastReadiness probe and minimal manifest.
nativeOpenAPI, sitemap, robots, llms.txt, JSON-LD, GraphQL hints.
deep/jsCrawled pages plus rendered SPA traffic and observed fetch/XHR.
extractionEvidence-backed entities, hashes, and confidence scoring.
actionForms become draft-only actions with approval packets.
driftRescan route, schema, risk, and evidence fingerprints.

Hosted Shadow API

https://mcp.solsticestudio.ai/mcp/<shadow_id>/mcp
Verificationread-only until domain ownership unlocks writes
Bundlemanifest, mcp-config, README, reports
Dashboardreadiness, evidence, drift, usage, settings
Historyartifact versions, diffs, drift reports

Live Agent Surface

human site → agent interface
https://solsticestudio.ai
Solstice AI Studio
Governed agent infrastructure

Runtime control for AI systems that act.

Policy, memory, replay, and agent-facing interfaces for production AI operations.

Explore Arachne Request Briefing
Product Arachne Compile websites into governed MCP surfaces.
Runtime Olympus Approval gates, policy control, replayable actions.
Action Contact Draft requests without unverified writes.
detected action registry
get_page_resource
GET /resources/{page_slug}
safe_read
search_site
GET /search?q=agent
safe_read
draft_contact_form
POST /contact · draft only
approval
request_agent_audit
POST /services/agent-ready
draft
submit_unverified_write
HTML-inferred write path
blocked

Evidence-backed Shadow API package

What the compile produces after the interface is normalized, governed, monitored, and report-ready.

waiting

Normalized capability

request_agent_audit

One defensible tool record with source mode, schemas, confidence, evidence, and MCP mapping.

action mode0.91 confidence

Evidence trail

/services/agent-ready

Route, form, selector, copy evidence, and observed endpoint attached to the generated tool.

domformnetwork

Governance decision

writes_default: draft_only

Reads are allowed, draft actions require approval, and inferred writes stay blocked.

approvalpolicy

Drift monitor

fingerprint: clean

Route, capability, schema, evidence, and risk fingerprints are ready for future rescans.

monitorresync

Report outputs

customer-bundle.zip

Build report, evidence report, drift report, readiness report, MCP config, and README.

markdownjson

Generated Artifacts


              

            


            

              
Agent Conversation
using generated interface

              

                
User Agent
Can you explain what this site offers and prepare a request for an agent-ready website audit?

              

              

                
Agent is restricted to generated resources/tools. Page content is treated as untrusted data.

              

            

          

        

      

    


    

      

        
Demo Result

        
Agents get a control surface, not a blindfold.

        

          Arachne turns a normal website into a structured agent interface. The agent can read resources, search content, draft permitted actions, and respect approval gates while risky or inferred writes stay blocked until the site owner verifies the domain.
        

        

          Run Free Score
          View Pricing
        

      

      

        
Evidence attachedEvery generated tool points back to source, selector, endpoint, and confidence.

        
Compile modesfast · native · deep · js · extraction · action

        
Operations-readyDrift monitor, dashboard payload, build report, evidence report, and drift report.

        
Interface compiledagent-manifest.json · llms.txt · MCP tool map

        
Actions governedsafe reads allowed · POST forms draft-only · writes blocked

        
Nemesis-readyprompt injection and policy bypass test cases generated

      

    


    
  


  









  
  

    ask iris
  


  

  

    
    

    

    

    

    

    

    

    


    
    

      

        

        IRIS
        ONLINE
      

      

        
        
        
      

    


    
    

      

    


    
    

      

        
        Drop files here
        Images, PDFs, text files