Aligna

Inspiration

Aligna began as a practical response to a problem I kept encountering while managing several very different projects with AI: the information existed, but understanding what actually deserved attention still required repeatedly reconstructing context.

The original Aligna system became a human-facing governance layer for projects, commitments, blockers and priorities. The WebMCP Challenge raised a more interesting question: what if an AI agent did not have to interpret that system indirectly through a visual interface?

WebMCP provided the missing bridge.

What it does

The challenge build turns Aligna into an agent-readable operating system while keeping the human in control.

It exposes five read-only WebMCP capabilities:

• Understand Aligna
• Review a project
• Identify blockers
• Review upcoming commitments
• Recommend the next area of attention

This allows an AI agent to interrogate the same structured operating state presented to the human, rather than trying to infer meaning from pixels, page structure or an unstructured conversation.

The result is a simple division of responsibility: Aligna holds the governed state, WebMCP exposes deliberate capabilities, the agent reasons over them, and the human retains authority.

How we built it

Aligna existed before the challenge as a personal life and project-management system. For the WebMCP Challenge, I created a separate public-safe competition build using fictional demonstration data and added the WebMCP interaction layer.

The competition work included defining and implementing the five WebMCP capabilities, creating the fictional public dataset, adapting the dashboard for a judge-facing demonstration, deploying the live application, publishing the open-source repository, and verifying the capabilities against the deployed build.

The WebMCP surface is deliberately read-only. The agent can understand, review and recommend, but it cannot silently alter canonical project state or assume human approval authority.

Challenges we ran into

The biggest challenge was not simply exposing data. It was deciding what an agent should be allowed to know and do.

A useful agent interface needs more structure than giving an AI unrestricted access to an application. Each capability therefore has a defined purpose and boundary.

A second challenge was demonstrating a real operating concept without exposing the private information contained in the operational Aligna system. The public competition build consequently uses entirely fictional, public-safe tracker data while preserving the structure and behaviour needed to demonstrate the concept.

Accomplishments that we're proud of

We turned an existing human-facing operating system into a working agent-readable application without giving the agent unrestricted authority.

The final build exposes five deliberate WebMCP capabilities, runs against a deployed public-safe dataset, and demonstrates those capabilities through a live application and open-source repository.

Just as importantly, the public competition build was designed as a separate, fictional-data environment so private operational Aligna data would not be exposed publicly.

What we learned

WebMCP changed how I think about application interfaces.

A dashboard can serve a human extremely well while remaining awkward for an AI agent. Giving the agent structured capabilities alongside the human interface creates a much cleaner model: humans and agents can interact with the same underlying system through interfaces designed for each of them.

That opens up possibilities well beyond Aligna. Any structured operational application could potentially expose a carefully governed agent interface without surrendering control of the underlying system.

What's next for Aligna

The next step is to explore governed write workflows in which an agent can propose changes while explicit human approval remains the authority boundary.

Longer term, Aligna could become an agent-native governance layer across multiple projects and workflows: not an autonomous system making decisions for the user, but a structured environment that helps humans and AI work from the same operational truth.

Built With

• react
• typescript
• webmcp