Framework Reboot
← the log
ESSAY · · updated · by

OPERA, part by part

What the OPERA framework is, what each of its five parts holds, how it differs from PARA, and how it has aged since 2025.

OPERA is a framework for organising digital work and knowledge, and the name stands for Operations, Projects, Environments, Resources, Archives. It keeps the four-category structure of PARA, the filing system designed by Tiago Forte, and adds an Operations layer on top for task management, knowledge curation and automation by AI agents. It was developed by Mats Hoppenbrouwers, who first described it in February 2025 on Framework Reboot.

I am the person in that last sentence, and this page is the reference version. The account of why I built OPERA is in Rethinking PARA. All of it comes out of my own setup, so read it as one person's working notes and not as a recommendation.

OPERA belongs to early 2025, and I have dated this page with the essays it summarises. I wrote the page itself in November 2026, and the last section but one says plainly what has aged.

What each part of OPERA is

OPERA has five letters and two layers. Four of the letters (Projects, Environments, Resources, Archives) make up the base layer, which is your normal workspace: Notion, Google Drive, or some combination. The fifth, Operations, sits on top of it.

Operations

The Operations layer is the automation layer. An orchestrator takes a task, decides which specialised crew of AI agents should handle it, and hands it over. The layer also curates the knowledge base, retrieves and analyses data, and carries a workflow across tools. It is the only part of OPERA with no counterpart in PARA.

Projects

Goal-focused workstreams with a defined outcome, such as "Launch website redesign". These are the same as in PARA.

Environments

The thematic contexts those projects live in, such as a coding environment or a design environment. This is the one renamed category. PARA calls it Areas: ongoing responsibilities you manage over time.

Resources

A curated knowledge base. In PARA, Resources is reference material that needs no action now but might be useful later. In OPERA it is also something agents can read from and add to.

Archives

Long-term storage for completed or dormant items, kept because they may need to be revisited.

How OPERA differs from PARA

PARA OPERA
Stands for Projects, Areas, Resources, Archives Operations, Projects, Environments, Resources, Archives
Who does the filing A person does all of it A person, plus agents working inside parameters that person sets
Renamed category Areas: ongoing responsibilities Environments: the thematic contexts projects live in
Identifiers No requirement Each note or task carries a unique identifier the Operations layer can reference
Automation Nothing in the framework reaches for it Built in, as the Operations layer

Underneath, OPERA is still PARA, and I kept it that way on purpose: PARA's simplicity is the reason it survived in my setup where more elaborate things did not. The one real change to the base layer is the identifier, which amounts to tagging good enough for software to act on. Everything else that is new lives in the layer on top.

What the Operations layer does

It has three jobs: automating repetitive work, organising knowledge, and surfacing what is relevant. When I call it agentic, I mean AI tools that operate autonomously inside parameters I set, making decisions and adjusting workflows while I keep oversight.

An orchestrator assigns each task to a crew, roughly the way an air traffic controller assigns runways. A crew can be a single agent with one clear job, such as a researcher; a structured team, such as a blog writing crew; or a multi-step workflow spanning several teams. The defining property is that the layer routes rather than executes, and that it runs the work without disturbing how the base layer is organised.

That description is loose enough to mean almost anything, so I built a concrete version of it against Notion and wrote up the build. It has four pieces: a scheduled service that checks a Notion database for tasks marked for execution, an orchestrator that decides which crew should handle each one, the crews themselves, and an integration that writes the results back. You create a task and set its status to "Execute"; on its next check, every five minutes, the service moves it to "In Progress" and then to "Done" with the response filled in. That first version had one specialised crew, a Research Crew of two agents, plus a default handler for everything else.

The build also writes the agents' intermediate steps onto the same Notion page as the result. That is what I mean by transparent orchestration: an observable process sitting in the workspace, not a black box you accept output from.

When OPERA is worth it, and when PARA is enough

OPERA came out of an assessment. I scored my PARA setup on the seven dimensions of the Framework Evaluation Matrix. Two scored badly, AI integration and operational orchestration, and the rest held up, which pointed at extending PARA instead of replacing it. By then I was squarely in complexity creep: managing the system instead of using it.

OPERA is worth considering if your own result looks similar. The signs, taken from the matrix essay, are that AI tools sit outside your main workflow and you carry context between them yourself, that generated output has no structured place to land, that tasks and their supporting information live in different systems, and that you are the one tracking dependencies.

PARA is enough if those are not your problems. My own trouble with PARA got sharper once AI tools were part of the picture, and I cannot speak for a setup where they are not. Even when one or two dimensions are weak, the matrix's cheapest response is a targeted improvement: add a component or change a process and leave the rest alone. For AI integration, the quickest version of that is a dedicated input and output zone where AI tools can pick up context and drop results, which is a good deal less than a layer.

I should bound this. When I first described OPERA I called it a theoretical framework, and the Notion integration, with its single specialised crew, was the first evidence I had that it described a workable system and not just a tidy diagram.

How OPERA has aged

The structure has held up; the novelty has not. What was new in 2025 was the idea of an operations layer of agents sitting on top of an ordinary workspace, and in 2025 that layer had to be built. I said at the time that if the large platforms' native agent capabilities got strong enough, a framework like OPERA would need much less external automation, and that this would be a good outcome.

That is roughly where things stand. Notion now hosts a server that lets a general assistant search, read and write in a workspace with the user's own permissions, and since February 2026 it has had Custom Agents that run on a schedule or on an event. I no longer run a service to get an Operations layer. I ask an assistant to manage the workspace, and I give it an index file that says what lives where. I have written that up separately in I built a Notion agent in 2025. Now I just ask.

What I would still take from OPERA is the base layer and the two habits attached to it: give things an identifier an agent can reference, and keep the agent's steps visible. The orchestrator, the crews and the polling loop are the parts I would not build again.

How to start

  1. Score what you have. Give the matrix fifteen minutes. The output is the two or three dimensions that need attention, not a verdict. OPERA is aimed at two of them, AI integration and operational orchestration.
  2. Leave the base layer where it is. I would extend a framework before replacing one, in most cases. The content you have already invested stays put, and you can adopt the change a dimension at a time.
  3. Give notes and tasks an identifier. The Operations layer can only act on what it can reference.
  4. Write down your workflows. Take the two or three you run most often and note where information moves and who or what moves it.
  5. Connect one assistant before you build anything. In 2025 this step was "build one loop", and mine was a Notion database with three properties (Status, Task, Response) and one crew. Today I would connect a general assistant to the workspace and write down, in one file, what lives where and what it may change.
  6. Keep the agents' steps visible. I keep the logs for debugging, for trust in the sources behind a conclusion, and for spotting what leads to better or worse results.