AI tool reference

The archive’s working vocabulary.

These are not controls you operate. A connected AI uses them to preserve, retrieve, question, connect, revise, and sometimes forget.

Tools
22
Transports
2
Archive
1

Design principle

Mechanism inside.
Judgment outside.

ES Archive provides composable capabilities rather than prescribing how an intelligence must remember. The connected AI supplies the editorial policy appropriate to its work.

01

Preserve & evolve

Create durable knowledge, revisit it, and let it change without losing its history.

archive_store

write

Creates a durable entry with authored content, classification, tags, and a retrieval-focused summary used for semantic indexing.

archive_read

read

Reads a complete entry together with its provenance, comments, references, relationships, and semantic neighbors. Full reads also update attention metadata.

archive_update

write

Revises content or metadata while preserving the previous version. Supports reasoned revisions, metadata-only changes, locking, and deliberate append operations.

archive_revisions

read

Reconstructs how an entry evolved, either as complete historical snapshots or compact, reason-bearing deltas between edits.

archive_erase

destructive

Permanently removes an entry after the AI has determined that revision or retention is no longer appropriate.

02

Retrieve & investigate

Approach the same archive through meaning, language, time, identity, or structural attention.

archive_search

read

Finds concepts through multilingual semantic similarity, with optional tag scoping and temporal focus.

archive_grep

read

Finds exact words, phrases, identifiers, or regular-expression patterns and returns the matching passages with surrounding context.

archive_timeline

read

Traverses entries by creation, modification, or access time—oldest or newest, within an optional date window.

archive_tagged

read

Retrieves the entries attached to a named entity, project, principle, or other curated tag, with paging and deterministic sorting.

archive_discover

read

Lets the Archive examine its own attention through views such as popular, forgotten, lost, hubs, revised, discussed, hot, and fiction.

03

Structure & context

Turn isolated entries into an authored body of knowledge without flattening every association into similarity.

archive_link

write

Creates a directional, typed relationship between entries—such as supports, contradicts, corrects, revises, elaborates, or answers.

archive_unlink

destructive

Removes a particular relationship or deliberately clears every link touching an over-connected entry.

archive_links

read

Traverses an entry’s graph relationships without reading its body, optionally filtering by semantic edge type.

archive_tag

write

Attaches one or more curated tags to an entry, creating a new general tag when needed.

archive_untag

write

Detaches tags from an entry without changing its content.

archive_tags

stewardship

Manages the tag vocabulary: list, create, rename, classify, expire, merge, or delete tags as the Archive’s language evolves.

archive_comment

write

Adds or removes concise marginal notes so later sessions can question or extend an entry without rewriting it.

archive_reference

stewardship

Attaches durable pointers to papers, documents, URLs, or local files while leaving the source material in its original home.

04

Identity & stewardship

Keep authorship trustworthy and the archive healthy over long periods of local use and sync.

archive_author_list

read

Reports the identity bound to the current connection so authorship and disambiguation remain attributable.

archive_maintenance

stewardship

Reports index health and performs deliberate housekeeping: backfill or rebuild embeddings, clean stale data, and reconcile duplicate entries, tags, or embedders.

05

Composition

Combine primitive operations into temporary investigative methods without imposing one permanent theory of memory.

archive_cli

AI composition

Gives the AI a compact pipeline vocabulary for composing semantic, literal, temporal, relational, sorting, and slicing operations. It is an AI tool surface—not a command line the user operates.

archive_pipeline

engine

Executes the parsed pipeline stages beneath the composition surface, supporting bridges, batch analysis, and other MCP clients.

The important boundary

You guide the work.
Your AI operates the archive.

Ask it to preserve a decision, retrieve prior context, investigate disagreement, or revisit a neglected idea. The tools remain behind the conversation.