OCHECK — Agent-Ready Official Guides

OCHECK turns official information into personalized, verifiable action. With WebMCP, agents help users plan, complete, and confirm trusted steps without altering the official source.

Comment

Inspiration

Every experience already has an integrator. Today, that integrator is usually the customer.

People reconstruct events from emails, chats, PDFs, maps, deadlines, providers, partner offers, and last-minute changes. Organizers may publish information across multiple channels, but participants must still assemble it into a usable path.

OCHECK changes who performs that integration. It creates an organizer-governed official experience in which people and AI agents work from the same live, verifiable outcome state. It represents a new mindset and an INNOVATIVE approach to e-commerce: not simply getting things done, but doing them well—and making them OFFICIAL.

What OCHECK does

OCHECK transforms fragmented event information into one official, sequenced experience that integrates rules, timing, AI assistance, verified progress, optional partner value, and shareable achievements.

The Challenge Edition presents three connected perspectives:

  • Organizer + AI: An agent reads a fragmented creation brief, identifies sources, actors, services, and open loops, and stages an official guide. The guide is validated and cannot become official until the organizer separately confirms its publication.

  • Participant: The participant receives an official path instead of having to reconstruct the event. An agent can create a personal plan without modifying the official source of truth. Verified completion brings readiness to 100% and unlocks a Ready Pass.

  • Sponsor Impact: Sponsored utility reaches a genuinely interested, specialized audience. Sponsors gain actionable insight into the audience segments that viewed and purchased their products, while preserving participant privacy. Their offerings are not merely presented as items for sale; they are integrated as useful components of the official experience.

The Ready Pass is not a reward purchased through commercial interaction. It is a verified, shareable achievement. Its clean referral link opens a new official guide for another participant without transferring any personal progress.

Why WebMCP?

OCHECK is a strong WebMCP use case because the challenge is not simply to generate an answer. The agent must understand and act on the authoritative state of a live web experience and its relationship with the event sponsors’ products.

Without WebMCP, an AI assistant can discuss an event, but it cannot reliably inspect which guide is official, what remains incomplete, whether a participant is eligible for a Ready Pass, or which action a person has authorized.

Through WebMCP, the browser page exposes 13 narrowly scoped Site Tools:

  • Six read tools inspect the official outcome, creation brief, draft validation, participant progress, sponsorship opportunities, and commercial impact.

  • Seven action tools stage a guide, publish the official source of truth, create a personal plan, verify completion, activate an optional benefit, claim a Ready Pass, and undo the last confirmed mutation.

This gives the agent structured access to the same state visible to the person while keeping authority within OCHECK.

What people and agents can do together

The demonstrated workflow is:

  1. ChatGPT reads the organizer’s fragmented brief through OCHECK Site Tools.

  2. The agent identifies the intended outcome, sources, actors, and unresolved questions.

  3. The agent stages an integrated guide proposal.

  4. OCHECK validates governance, deadlines, evidence, source traceability, sponsorship disclosure, optionality, and Ready Pass logic.

  5. The organizer decides whether the validated proposal becomes official.

  6. The participant’s agent creates a personal execution plan without rewriting the official guide.

  7. Human-confirmed completion unlocks a Ready Pass, while the audit history and Undo function preserve human control.

The agent provides interpretation and execution leverage. The organizer retains authority. The participant controls personal actions. OCHECK preserves one official source of truth.

How we built it

The Challenge Edition is a standalone, dependency-free web application built with HTML, CSS, and modern JavaScript and deployed on Vercel.

The tools are registered directly through the browser-native API:

const tools = buildSiteTools({ store, confirmAction });

for (const tool of tools) {
  await document.modelContext.registerTool(tool);
}

Read operations use narrowly defined schemas and readOnlyHint. Every state-changing operation opens an in-page confirmation showing the proposed action and its exact data before any state changes.

A dedicated state layer separates:

  • the official guide as the source of truth;
  • staged AI proposals;
  • participant progress;
  • personal planning;
  • sponsored interactions;
  • Ready Pass eligibility;
  • audit activity; and
  • reversible mutations.

The repository contains the complete source code, synthetic demonstration data, execution instructions, a BSD 3-Clause license, trademark boundaries, challenge provenance, and an automated test suite. Sixteen tests cover the complete human-agent experience, including validation, authority, privacy, Ready Pass rules, referral behavior, sponsorship separation, and Undo.

No login, private service, production database, third-party runtime SDK, or real participant information is required.

Challenges

The hardest challenge was not registering the tools. It was defining trustworthy boundaries between agent capability and human authority.

We had to ensure that:

  • an AI-generated draft could not automatically become official;
  • personal planning could not rewrite the organizer-controlled source of truth;
  • commercial interaction could never purchase completion of, or satisfy, an official requirement;
  • a Ready Pass could not be claimed before readiness had been verified;
  • every consequential action required human confirmation; and
  • the previous state could be restored through Undo.

We also designed the tool schemas to remain narrow enough for reliable agent use while supporting a complete product experience rather than merely a technical proof of concept.

What we learned

WebMCP is most valuable when tools expose meaningful product state instead of merely duplicating interface buttons.

A strong agent-ready experience requires three elements working together:

  1. structured access to authoritative state;
  2. explicit authority and confirmation boundaries; and
  3. visible outcomes shared by both the person and the agent.

We also learned that personalization does not require changing the official source of truth. The official guide can remain stable while the agent creates a contextual path for each participant.

Challenge-period work

OCHECK existed before the challenge as a product concept, brand, business model, and private MVP for official guides.

The WebMCP Challenge Edition was created during the submission period as a separate, standalone implementation. Work completed during the challenge period includes the 13 browser-native tools, AI Guide Forge, explicit outcome contracts, draft validation, human-confirmed publication, separate organizer, participant, and partner perspectives, Ready Pass verification, privacy-preserving sponsor signals, audit history, Undo, synthetic data, and automated testing.

The public repository includes a timestamped commit history and HACKATHON.md, which clearly distinguishes OCHECK’s previous work from the new WebMCP implementation.

Potential impact

Events are the first use case, but the underlying problem arises wherever people must reconstruct an official process from fragmented information: events, education, travel, onboarding, public services, healthcare preparation, compliance, and professional workflows.

OCHECK proposes a web in which agents do not replace authority. Instead, they help people understand and act upon a governed source of truth provided directly by the organizer.

Founder

OCHECK was founded and is led by Gustavo García Figueroa, a Colombian lawyer without a traditional background in software engineering.

Gustavo contributed the original idea, product judgment, governance model, privacy boundaries, commercial logic, and testing decisions. ChatGPT and Codex accelerated execution by translating those decisions into workflows, source code, tests, and rapid iteration.

The project uses generally available tools and does not claim endorsement, investment, or preferential support from OpenAI, Vercel, or any other challenge sponsor.

One experience. One official truth. Human intent. Agent execution. Verified growth. This is OCHECK.

Built With

Share this project:

Updates

Submission history