HermesMarkdown is a specialized Markdown editor designed for users who need their notes to be easily readable by AI agents. It belongs to the category of productivity tools for AI-assisted workflows, targeting developers, technical writers, and project managers who rely on tools like Claude Code or Cowork. The core value of HermesMarkdown is that it structures your plain .md files so agents can identify relevant content without scanning everything, preserving precious context window space. By generating an AGENTS.md index and scoring the readability of each file, the editor ensures that agents query notes like a database rather than brute-forcing through entire vaults. This approach eliminates the need for manual tagging or external databases, keeping everything in local Markdown files.
The concrete problem HermesMarkdown solves is the inefficiency of AI agents when dropped into a vault of unstructured notes. Without prior knowledge, agents start from scratch every time, burning through context windows as they read irrelevant files or guessing wrong and pulling incorrect data. This leads to wasted tokens, slower responses, and frustrating manual file wrangling. For users who regularly ask agents to find specific information across hundreds of notes, the overhead becomes a significant bottleneck. HermesMarkdown addresses this by providing a protocol—the three-tier index in AGENTS.md—that tells agents at a glance which files to scan, which to read, and which to load fully. This systematic filtering cuts token usage and retrieval time dramatically.
The first major feature group is the AGENTS.md index with its three-tier protocol. When a vault is opened, HermesMarkdown automatically generates an AGENTS.md file that defines how agents should approach each note. The tiers are: scan read_when to filter without loading files, read scope to confirm relevance, and load the full file only when necessary. This hierarchy ensures that an agent reads only the metadata first, making quick decisions without waste. The status bar displays a live readability score for the current file, drawn from frontmatter completeness, heading structure, typed code fences, table usage, and bold anchors. Users can see at a glance whether a file is agent-ready, without opening separate tools. This feature directly reduces token consumption and speeds up interactions.
The second major feature group is the frontmatter wizard and customizable metadata fields. HermesMarkdown comes with four default fields—scope, read_when, related, and tags—but users can drop, rename, or skip any of them entirely. A wizard walks users through these fields when creating new files, making setup painless even for non-experts. The editor also scores heading structure, typed code fences, table formatting, and bold anchors, so even a plain note without frontmatter earns a meaningful readability rating. This flexibility means that users can adopt as much or as little structure as their workflow demands. Every agent that reads the vault benefits from whatever structure is added, and the improvements are permanent. The wizard ensures consistency without forcing a rigid schema.
admin
The third feature group includes BYO AI keys, Smart Workspaces, and WikiLinks. Users connect their own Anthropic or Google Gemini API key to unlock AI assistance directly in the editor—auto-generating scope fields, suggesting related links, and improving writing inline. Keys are stored locally in the browser; they never reach HermesMarkdown’s servers. Smart Workspaces allow filtering the entire vault by tag, date, or word count, with a dedicated Today’s Work view showing items touched in the last 24 hours. WikiLinks ([[note name]]) connect notes vault-wide, and Ctrl+Click navigates instantly. These capabilities make the editor both a standalone writing tool and an AI-powered assistant that respects privacy. The local-first approach means no uploads are required.
Overall, HermesMarkdown operates on a local-first, browser-based workflow. Users open any local directory as a vault and save directly to their machine without creating an account. The editor generates AGENTS.md automatically, and a readability score is always visible in the status bar. Optional Google Drive sync provides background backup without moving files off the local system—disconnect anytime and files remain. For power users, Zen Mode (Ctrl+Shift+Z) collapses panels, multi-pane editing allows side-by-side file comparisons, and live elements like clickable checkboxes, date pickers, and WikiLink navigation make editing interactive. Shortcodes expand inline: ..d for today’s date, calc() for expressions, and / for template menus (Daily Note, Meeting Notes, etc.). The entire experience is designed to keep control with the user.
Concrete use cases illustrate the outcomes. A developer using Claude Code to query project plans opens a vault with dozens of notes; instead of loading everything, the agent reads AGENTS.md and uses read_when to locate only the files tagged with “v1 launch plan”. Result: quick, accurate retrieval with minimal token usage. A project manager tracking ops logs adds a Total: line to a list of expenses, and HermesMarkdown sums them automatically—no spreadsheet needed. A writer organizes ideas across multiple files by tagging them in frontmatter and linking via WikiLinks; the Today’s Work view shows exactly what was edited recently. For backup, Google Drive sync ensures files are safe without vendor lock-in. These scenarios yield faster query responses, reduced context waste, and simpler workflows.
HermesMarkdown targets developers, technical writers, and project managers who regularly interact with AI coding agents like Claude Code, Cowork, or similar assistants. It runs in any modern browser and saves files locally, requiring no account or cloud dependency. The tech stack is plain Markdown—no proprietary formats—and the AGENTS.md protocol is a convention that any agent can follow. Pricing is free, with no hidden tiers; users only pay for their own API keys if they choose to use AI features. In summary, HermesMarkdown delivers a structured, agent-friendly Markdown environment that protects context windows and accelerates retrieval, all while keeping data under the user’s control. It transforms a simple plain-text vault into a queryable database for AI.
Developers using AI coding agents such as Claude Code and Cowork, technical writers managing documentation vaults, project managers tracking budgets and tasks in Markdown, and power users who want their notes to be easily queried by AI assistants without excessive token consumption. This tool is also suited for anyone who maintains a local Markdown vault and wishes to make it agent-friendly without moving to a proprietary platform.