AIエージェントが直接書き込むノートアプリ「hanamask」を作ったhanamask v1.0.0 is a desktop note and task management app designed for AI-agent…

A developer has released hanamask v1.0.0, a desktop note and task management application built specifically for working alongside AI agents. The project, announced on the Qiita developer blog, is notable because it inverts a common assumption about note-taking software: rather than treating the human as the primary author, hanamask is designed around the idea that the AI agent writes the content and the human reads and reviews it. That design choice matters as more developers integrate AI assistants into daily workflows and look for tools that can serve as durable, shared memory between people and machines.

The concept, as the author summarizes it, is that "the body is written by the agent, and the human reads it." In practice this means an AI agent can compose and update notes directly, producing Markdown-formatted content that a person can then scan, verify, and act on. The app positions itself as optimized for AI development and human-agent collaboration, which suggests the intended audience is programmers and other technical users who already spend significant time prompting and supervising AI tools while building software.

The technical mechanism that makes this possible is the Model Context Protocol, or MCP, which the tags and category identify as central to hanamask. MCP is an open standard, originally introduced by Anthropic, that defines how AI applications connect to external tools, data sources, and services. An MCP server exposes capabilities—such as reading, creating, or editing notes—that a compatible AI client can call. By implementing this interface, hanamask appears to let an agent write into the app's note store without a human manually copying and pasting text, so the notes become a live workspace that the agent can modify as a task progresses.

Markdown is the other key ingredient. Because the content is stored as Markdown, notes remain lightweight, human-readable, and portable, and they render consistently across many editors and platforms. Markdown is also well suited to output from large language models, which tend to produce clean, structured text with headings, lists, and code blocks. Combining Markdown with task management features implies that hanamask can hold not just free-form notes but also actionable items that an agent updates over time, though the exact scope of the task features is not detailed in the announcement.

To understand why an app like this is emerging now, it helps to look at the broader ecosystem. Over the past year, MCP has gained rapid adoption as a way to standardize the connections between AI models and the tools they use, and a growing catalog of MCP servers now covers everything from file systems and databases to issue trackers and design tools. Major AI clients and coding assistants have added support for the protocol, which reduces the need for each application to build bespoke integrations. hanamask fits into this trend as a purpose-built endpoint: instead of bolting AI onto an existing notes product, it treats agent access as a first-class feature from the ground up.

Adjacent tools help frame where hanamask sits. Popular note applications such as Obsidian and Notion have added AI features or community plugins, and Obsidian in particular relies heavily on local Markdown files, a model that resonates with hanamask's approach. General-purpose task managers and developer-focused knowledge bases are likewise experimenting with agent integrations. What distinguishes hanamask, based on the announcement, is the explicit framing that the agent is the writer and the human the reader, rather than the AI acting only as an occasional assistant to a human author.

As with any early release, some caveats apply. This is version 1.0.0 from what appears to be an individual developer, so the feature set, stability, and platform support are likely still maturing, and the announcement does not spell out pricing, licensing, or which operating systems are supported. Users who let an agent write directly into their notes will also want to consider review and verification practices, since the value of the "human reads" half of the concept depends on catching errors or hallucinated content. Still, hanamask is a concrete example of software being redesigned around AI agents as active participants, and it illustrates how MCP is enabling that shift in everyday productivity tools.