Skip to content

// MidPledge

The Trust Layer for
AI Delegated Authority

Platforms govern agents. MidPledge assures their authority to the outside world.

When an AI agent acts across an organizational boundary, the other side needs to know what it is authorized to do. MidPledge makes that authority current, transaction-bound and independently verifiable.

See how it works
cross-boundary authority
Cross-boundary authority, illustratedCompany A, the principal, authorizes an authority source; its agent, Agent-27, sends an intended action across the trust boundary. MidPledge resolves and checks the agent's current authority at that source and binds it to the exact transaction. Organization B, the relying party, independently verifies, makes its own decision, and both sides receive an Authority Receipt. Illustrative diagram using sample data.COMPANY A · PRINCIPALORGANIZATION B · RELYING PARTYTRUST BOUNDARYAuthority sourcepolicy · grants · v12Agent-27acts for Company AMidPledgeresolve · check · bindIndependent VerifyAUTHORIZED · MATCHOwn decisionACCEPT under own policyAuthority ReceiptMP-TX-DEMO-7C21Cross-boundary authority, illustratedCompany A, the principal, authorizes an authority source; its agent, Agent-27, sends an intended action across the trust boundary. MidPledge resolves and checks the agent's current authority at that source and binds it to the exact transaction. Organization B, the relying party, independently verifies, makes its own decision, and both sides receive an Authority Receipt. Illustrative diagram using sample data.COMPANY A · PRINCIPALORGANIZATION B · RELYING PARTYTRUST BOUNDARYAuthority sourcepolicy · grants · v12Agent-27acts for Company AMidPledgeresolve · check · bindIndependent VerifyAUTHORIZED · MATCHOwn decisionACCEPT under own policyAuthority ReceiptMP-TX-DEMO-7C21
Illustrative diagram using sample data.

// The trust gap

Internal governance isn't external assurance.

Internal platforms govern their own agents — policies, permissions, approvals. An external relying party can't see any of that. For a specific action, it needs evidence of what the agent is authorized to do that it can check for itself.

Verify authority at the source — not from the agent's claim.

The relying party's question

“Is this agent genuinely authorized by the organization it represents to perform this specific action, under these limits, right now?”

  • Inside: policies · permissions · approvals
  • Across: source-resolved, current, transaction-bound evidence

// Authority in action

One transaction, across the boundary

Principal → agent and authority source → transaction binding → MidPledge verification → relying-party decision → Authority Receipt. Step through it, then change the transaction yourself.

Journey diagram — the highlighted parts follow the selected stageAuthority source, Agent-27 and Company A on one side of a trust boundary; Organization B, Verify and decision on the other; MidPledge in the middle; the Authority Receipt at the end. The stage panel describes the highlighted part.COMPANY A · PRINCIPALORGANIZATION B · RELYING PARTYTRUST BOUNDARYAuthority sourcepolicy · grants · v12Agent-27acts for Company AMidPledgeresolve · check · bindIndependent VerifyAUTHORIZED · MATCHOwn decisionACCEPT under own policyAuthority ReceiptMP-TX-DEMO-7C21Journey diagram — the highlighted parts follow the selected stageAuthority source, Agent-27 and Company A on one side of a trust boundary; Organization B, Verify and decision on the other; MidPledge in the middle; the Authority Receipt at the end. The stage panel describes the highlighted part.COMPANY A · PRINCIPALORGANIZATION B · RELYING PARTYTRUST BOUNDARYAuthority sourcepolicy · grants · v12Agent-27acts for Company AMidPledgeresolve · check · bindIndependent VerifyAUTHORIZED · MATCHOwn decisionACCEPT under own policyAuthority ReceiptMP-TX-DEMO-7C21

Interactive example · Sample data · a presentation sequence, not a protocol specification.

// Verify Explorer

Change the transaction. Watch the decision.

One worked example — an agent places an order with an external supplier. Move the limit, swap the counterparty, revoke authority or take the source offline.

verify · explorer Agent-27 · Company A → Supplier B
Agent claimsignored

“authorized up to USD 100,000”

Authority below comes from Company A's authorized source: limit 10,000 · approval above 7,500 · approved suppliers only.

6,000 USD
500approval > 7,500limit 10,00015,000

Supplier

Authority state

Human approval

Authority source

resolve authority bind rp verify rp decision execute record
agent creates transaction

          

          

            
supplier b verifiesAuthorized

            

          

        

      

      
Interactive example · Sample data  runs entirely in your browser; nothing is sent to MidPledge. Endpoints and fields are examples. A source outage returns UNVERIFIABLE, never an optimistic authorization.

    

  

  


  
  

    

      

        
// Applications

        
Different actions. One authority problem.

        
The same reusable check applies wherever an agent's action has consequences for someone outside its own platform.

      

      

        
41use cases mapped

        
8authority families

      

    

    

  
B1 · Enterprise System & API

  
Production API Action

  
Agent invokes a consequential external or internal API under action/resource limits.


  
C2 · Data

  
External Data Sharing

  
Agent releases data to another organization under purpose/recipient constraints.


  
F3 · Delegation & Multi-Agent

  
Cross-Company Agent Delegation

  
One organization authorizes another party's agent for a bounded task.



    

      
A map of where the model applies — not a list of currently available products.

      Explore all 41 use cases
    

  


  
  

    

      
// Who MidPledge serves

      
Two sides of one trust boundary

      
The organization that delegates authority and the organization that relies on it need different things. The trust checkpoint connects them.

    

    

      

        Authority issuers / Principals
        
“Our agents act for us. Others need to trust that.”

        
    
          
  • Keep governance in the sources you authorize
    • 
          
  • Set scope, limits and approvals once
    • 
          
  • Revocation takes effect at the next verification
    • 
        

      

      

        
        
MidPledge

        
trust checkpoint

        
resolve · bind · verify

      

      

        External relying parties / Counterparties
        
“An agent wants us to act. Is it authorized?”

        
    
          
  • Verify this exact action yourself, before accepting
    • 
          
  • See current state — active, revoked or unverifiable
    • 
          
  • Decide under your own policy; keep the evidence
    • 
        

      

    

    
Platforms, networks, trust providers and developers participate on either side. Roles and categories — not customers or partners. For enterprises →

  


  
  

    

      
Start a conversation

      
Whether you delegate authority to agents, receive their actions, build them or want to explore a partnership.

      Register Interest