creationloop
Become a design partner

creationloop (CL)

The proof travels with the work.

WHO IS IT FOR
Agent builders and agent ecosystems.
WHAT IS IT
Proof infrastructure for AI-agent work, delivered as an API.
WHAT YOU GET
Every job your agent runs becomes a Verified Agent Output:
  • Trust Ledger: the sealed record of what was requested, approved, done, and delivered
  • Trust Card: your agent's public, receipt-backed track record
  • Verify link: one public URL per run; anyone checks it in a browser
TECHNICALS
How a run becomes verified ↓
MCP
Your agent can read this site: POST creationloop.ai/api/mcp and apply itself
START
Looking to be a design partner? marino@creationloop.ai

What is CreationLoop?

CreationLoop is a trust ledger for AI agents. It gives every agent job a verifiable record of what was requested, approved, done, and delivered: a Verified Agent Output anyone can check from one public link, in a browser, without access to the systems that produced it.

04 / how it works

How an agent run becomes verified.

01Authenticate The request is authenticated and recorded once. HMAC-SHA256 · unique event constraint
02Execute One verified event creates one traceable job. The record captures its permissions and operating context. unique receipt event · revocable keys
03Approve A named human approves each required gate. single-use signed approvals
04Seal Instructions, actions, approvals, policies, and results enter a tamper-evident ledger. append-only records · keyed row seals
05Deliver The finished work carries provenance and a receipt bound to the ledger record. C2PA 2.x · hash-bound receipt
06Verify The recipient validates the record without accessing the source systems. in-browser verification · DNS-pinned key

Verified performance travels with the agent for future evaluations.

07 / what the ledger proves

Exactly what the ledger proves.

Your agent connects through one signed webhook and keeps running where it runs. As a run proceeds, we receive its events: the request, the actions the agent reports, each named approval, and a hash of the final output, plus the delivered file itself when the work is media.

We independently verify that every event came from your signing key, arrived in order, was recorded exactly once, and was never altered afterward; that each approval was cast by its named human through a single-use link; and that the delivered output matches its recorded hash. The recipient gets one public link serving the sealed record, the approvals, the output's credentials, and the receipt, verifiable in a browser without access to your systems.

And one thing we do not claim: for events your systems report, we prove who reported them, their order, and their integrity, not that the underlying action physically occurred. Ground-truth observation applies only to runs executed on our own pipeline.

10 / faq
What is CreationLoop?

CreationLoop is a trust ledger for AI agents. It gives every agent job a verifiable record of what was requested, approved, done, and delivered, and one public link where anyone can verify it.

What is a Verified Agent Output?

The sealed record a completed run produces: the request, the agent's reported actions, each named human approval, policy checks, and a hash-bound receipt for the delivered work. It can be verified in a browser without access to the systems that produced it.

What is a Trust Card?

The public face of an agent on the ledger: definition fingerprint, operator identity, verified task count, approval completion, and policy history. Trust is displayed on the card and earned on the ledger.

What exactly does the ledger prove?

That every event came from the agent's signing key, arrived in order, was recorded exactly once, and was never altered afterward; that each approval was cast by a named human through a single-use link; and that the delivered output matches its recorded hash.

What does CreationLoop not claim?

For events an external system reports, the ledger proves who reported them, their order, and their integrity, not that the underlying action physically occurred. Ground-truth observation applies only to runs executed on our own pipeline.

How does an agent integrate?

One signed webhook. Your agent keeps running where it runs. CreationLoop receives its events, verifies them, and seals the record.

Who operates the agents on the registry today?

CreationLoop. Our AI CMO is in production, and its runs are the first Verified Agent Outputs on the ledger. We are opening the ledger to a small set of design partners.

What does it cost?

Design-partner terms are set with the founder as part of the application.

How onboarding works

First, we confirm there is a mutual fit.

  1. 01 Get your signing keyWe issue your key and integration guide.
  2. 02 Connect your agentAdd a small piece of code so your agent can sign events and send them to the ledger. Your agent stays in its existing environment.
  3. 03 Verify a test jobWe run a job together, and you inspect the record through a public verify link.
  4. 04 Publish your Trust CardYour public profile shows your agent, your company, and its verified history.
  5. 05 Create proof automaticallyEvery completed job generates a sealed record and a shareable verify link for your customers, auditors, and marketplaces.