TrailMark
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| get_contextA | What the person is looking at in TrailMark right now: the open document (path relative to the project), whether they are reading or editing it, the section on screen (heading path and line range), the text they selected (with its section and lines) and whether it has unsaved edits. Call this when the person says "this", "here", "the selected text", "this section", "what I'm reading" or otherwise points at something on their screen without naming it. Needs TrailMark open on the project with its agent connection on; otherwise it says so. |
| list_commentsA | The comments the person left in TrailMark on passages of the project's Markdown documents ("this contradicts section 2", "simplify"). Each has an id, the document path, the quoted passage, its heading path and line, the comment text, when it was made, and |
| list_notesA | The notes the person chose to share with agents in TrailMark (its Notes window keeps their quick notes, drafts and prompts, and nothing in it is visible to you unless they shared that note). Each has |
| list_tagsA | The tags in use in the project's Markdown documents: each tag with how many documents have it ( |
| list_unreadA | The Markdown documents that changed since the person last read them in TrailMark, newest change first, with the time of the change. Call this when the person asks what changed, what they have not read yet, or to catch up (for example after you or another agent edited documents). Also works when TrailMark is closed, from the project's files. |
| openA | Show the person a place in TrailMark: a document, optionally at a line or a heading. TrailMark opens it in its window without taking focus from your terminal, highlights the section and shows a notice with a way back; the person looks when ready. Use it after you wrote or found something they should see ("see the new section in docs/plan.md"). It starts TrailMark on the project if it is closed. It only shows; it never changes a document. |
| read_noteA | Read one of the notes the person shared with agents, by the |
| resolve_commentA | Mark one comment done after you addressed it, with an optional short |
| searchA | Find where something is described in the project's Markdown documents. Returns the best matching sections ranked, each with its path, heading path, line and a short snippet (the same index as TrailMark's Search everywhere), and the |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 9 tools
Each tool targets a distinct resource or action: get_context (live screen state), list_unread (changed docs), search (locate sections), open (show the person), list_comments/resolve_comment (comment lifecycle), list_notes/read_note (note access). The cluster of read-only listing tools (list_comments, list_notes, list_tags, list_unread) shares a similar 'what does the person have' flavor and could momentarily be confused, but descriptions clearly separate them.
Most tools follow a predictable verb_noun pattern (get_context, list_comments, list_notes, list_tags, list_unread, read_note, resolve_comment), all snake_case. Two tools—open and search—are bare verbs without an object, which is a minor deviation but still natural and readable, especially as both are unambiguous.
Nine tools is well within the ideal 3-15 range and each earns its place, mapping the observable surface of a documentation project (context, comments, notes, tags, unread, search, display). No redundant or filler tools.
The surface covers the person's annotations and project awareness well: comments (list + resolve), notes (list + read), tags, unread tracking, search, and display. Notable gaps remain—there is no tool to return raw document content (search gives only snippets, open only displays), and no way to create comments or edit documents, though the latter is an explicit design constraint ('the only change an agent can make').