MCP reference

When connected via MCP, your AI client has access to 26 tools for working with your Nessie library and usage data. The filesystem-shaped coreutil tools are the canonical interface.

Discovery and analytics

Tool	What it does
`nessie_team_list`	List teams and team-shared resources readable by the authenticated user
`nessie_integration_list`	List readable integration and directly shared roots with access-path and owner provenance
`nessie_check_in`	Return generated profile sections and recent Nessie activity for session startup
`nessie_who_am_i`	Read your profile sections (biography, expertise, projects, decisions, etc.)
`nessie_analytics`	Load token and request usage totals, breakdowns, and trends for you or a team you administer
`nessie_skill_analytics_overview`	Load the all-skills analytics overview: invocations, unique people, success rates, and trends per skill for you or a team you administer
`nessie_skill_analytics`	Load one skill's dashboard, including evaluated recent invocations with their originating session IDs

Filesystem-shaped reading and navigation

Tool	What it does
`nessie_ls`	Browse available source worlds (contexts, transcripts, profile, Obsidian vaults) and traverse their hierarchy. A `shared` column flags outgoing grants on integration roots, folders, and contexts; directly shared conversations and agent sessions do not expose that column, so its absence does not prove a session is private. Incoming shares appear under the `owner` column instead
`nessie_grep`	Search readable content with hybrid or literal matching, parent/owner/repo scoping, raw kind filters, and time windows
`nessie_cat`	Read a node's full content as text
`nessie_head` / `nessie_tail`	Read the beginning or end of long content without loading all of it
`nessie_stat`	Inspect node metadata without reading its body
`nessie_asset_get`	Inspect an image asset referenced by a context

Current writing tools

Tool	What it does
`nessie_mkdir`	Create a folder, optionally inside an Editor shared folder
`nessie_tee`	Create a context with Markdown, emoji, folder placement, and source provenance
`nessie_replace_lines`	Safely replace or delete complete lines, including table rows and list items
`nessie_sed`	Replace an exact inline string in a context
`nessie_mv`	Move, rename, unfile, or change the emoji of a context
`nessie_rm`	Delete a context
`nessie_update_profile_card`	Update fields on a structured profile card
`nessie_rename_folder`	Rename a folder and optionally set or clear its emoji
`nessie_move_folder`	Move a folder into another folder or back to the top level
`nessie_rmdir`	Delete an empty folder
`nessie_set_slice_modality`	Set or clear an explicitly confirmed, owner-only correction on a cloud search result
`nessie_delete_conversation`	Delete a synced conversation or transcript and exclude it from future syncs

nessie_delete_conversation is two-step: call it first without confirm to get a preview (the conversation title and message count), show that to the user, then call again with confirm: true to delete. It is for conversations and transcripts only; nessie_rm handles contexts, and the two stay separate.

MCP access grants full read and write access to your Nessie library under the nessie:full scope.

Context Markdown, title, and emoji are synchronized collaborative fields. Owners and Editors can read and update the same current document through MCP while people work in the app; Viewers can only read. Editor shared folders accept caller-owned contexts and folders as new children, but moving them does not change ownership. MCP does not create or revoke sharing grants.

Resume / takeover

Resume is a workflow over search and read, not a separate MCP tool. If you paste a Nessie UUID for an agent transcript, node, source, or conversation into an MCP client, it should call nessie_head with lines: 5, then nessie_tail with lines: 40. The head window recovers the original goal and constraints; the tail window recovers the current state, latest decisions, blockers, and next actions.

If you do not have an ID, search transcript sources first:

Search transcripts for the project, branch, error, file path, or topic.
Choose the matching result ID.
Read its beginning with nessie_head and its recent end with nessie_tail.
Write a compact takeover brief, then continue the work.