Notes that execute: agent blocks and Notes → Actions in Navin

In Notion, a page describes work. In Navin, a note can launch it. That is the promise of agent blocks and the AI memory in the Notes module: the document is no longer a report, it is a command post.

This article shows how to go from "notes that document" to "notes that execute", with concrete examples.

The glass ceiling of note-taking tools

Every modern notes tool can format, link and organize. Some add an AI that summarizes or rewrites. But none crosses the line: acting. Writing "scrape the 50 largest AI integrators" in Notion produces... a nice page containing that sentence.

Navin is an agent platform first: terminal, files, browser, scraping, git, document generation. The Notes module lives in the same place as those tools - they just had to be wired together.

The /agent block

In any note, type /agent. A block is inserted with a structured template:

Objective:
Find the 50 largest AI integrators in Saudi Arabia

Sources:
Web + internal documents

Output:
Table with company, size, contact

The block renders as a card with a Run button. On click, Navin builds a structured prompt - the block spec, the note title and its file path - and places it in the chat composer. The agent executes (web search, scraping, file reading, code...) then appends the result to the note, under a ## Result section.

The block is a plain Markdown fenced code block with the agent language: it survives exports and git commits, and the agent can re-read it with its file tools. No magic format.

Notes → Actions: three real scenarios

1. From spec to code

A "Authentication bug" note lists the affected files and the error. Ask Navin with "fix this": the agent reads the note, opens the files, fixes, runs the tests. The note documents AND triggers the fix. With the Code module, the same gesture covers planning, implementation and review.

2. From meeting to tasks

Your meeting notes contain - [ ] items. The module's Tasks section aggregates them all, toggleable, with priorities. And the agent can turn minutes into an assigned action list - it reads the note, it writes the tasks.

3. From market study to leads

A "market study" /agent block with web sources: Run, and the agent scrapes, structures, writes the results table into the note. The note becomes a living deliverable, not a placeholder.

The memory that powers everything

What makes these scenarios reliable is that the agent knows your notes. They are semantically indexed in the background (local embeddings by default, via Ollama), and the agent's notes tool runs a hybrid semantic + lexical search.

Result: "What did we decide on pricing in July?" does not start from scratch - the agent finds the decision note, quotes the passage, and can chain into action. Your notes stop being an idea graveyard: they are your agents' working memory.

Why not in Notion?

Because execution requires what Navin already has and a cloud notes tool never will: a terminal on your machine, your git repos, your files, your API keys, your local models. Notion's AI lives in the Notion workspace. Navin's agents live at your place, next to your notes - in Markdown files you own (see the full comparison).

Getting started

1. Open Notes in the Navin sidebar.
2. Create a note, type /agent, describe the objective.
3. Click Run, review the prompt, send.
4. The result lands in the note. Iterate.

The module's Commands section lists every available block and tip (wikilinks, drag & drop, shortcuts). The rest is your imagination.