Empirical

Poll the status of an install. Use this if install_blueprint returned jobStatus "running" (still installing past its own bounded wait).

get_blueprint_install_status

Get current workspace selection for this OAuth client. When the selected team workspace has an active shape, also returns shape (its kinds of item with their keys, statuses and fields) and policy (the rules its managers confirmed), so use them when saving to or reading from that workspace.

get_current_workspace

Read the authenticated user's normalized Empirical behavior policy. This is read-only and reflects the current installer/settings preferences.

get_empirical_policy

Get one item and every item nested under it, at any depth, as a readable markdown document plus the list of nodes. Use this instead of walking the graph with get_memory_neighbors when you want a whole idea with its sub-ideas: it returns the complete subtree in one call, in the order the tree is arranged. Items nest only in a workspace that has a shape; for an ordinary memory this returns just that memory.

get_memory_tree

Retrieve a specific memory by ID. Returns the full summary and notes; use it when a query result shows summaryTruncated.

get_memory

Read one file attached to a memory. A text file (html, markdown, json, csv, code, plain text) comes back in full in the result, up to maxInlineBytes (default 512KB, at most 1MB). An image comes back as an image. Anything else, or anything larger, comes back as metadata plus a download link that works for ten minutes and a resource URI you can read with resources/read. Find the attachmentId with list_memory_attachments. Identify the memory by memoryId, or by a query that resolves to exactly one memory.

get_memory_attachment

Get related memories and edges for a memory node (incoming, outgoing, or both).

get_memory_neighbors

Get the authenticated Empirical profile/account context for the connected user. Not a source for weather, web facts, or general knowledge. When Empirical Tokens are enforced for this account, includes empiricalTokens: Empirical Tokens left this month, the monthly grant and the reset date.

get_user

Install a blueprint into your memory graph. Automatically claims first if the blueprint is free and not yet owned. Waits for the install to finish (up to ~20s) and returns the final status. A blueprint that uses Empirical Tokens first returns a cost preview; tell the person the cost, and call again with acceptEmpiricalTokenCost true only after they agree.

install_blueprint

Browse the blueprint catalog. Returns a compact summary per blueprint (not full detail — use get_blueprint for that). Optionally filter by category.

list_blueprints

List available guided-interview blueprint setups. Locked entries require a paid plan.

list_guided_setups

List every blueprint currently installed for you, with node/relationship counts and whether an update is available.

list_installed_blueprints

List memories to review and pick targets for follow-up actions. Use this for recent/latest/today requests; follow nextCursor while hasMore is true.

list_memories

List the files attached to a memory, with filename, size, type and upload date. Identify the memory by memoryId, or by a query that resolves to exactly one memory. Set includeDeleted to also see files that were deleted but are still inside the 30-day restore window, each with the date its bytes get purged.

list_memory_attachments

List selectable workspaces for this OAuth client (Default Workspace first).

list_workspaces

Read the memory usage policy for when to create, query, update, or delete memories.

memory_usage_guide

Preview what installing a blueprint would create: node/relationship counts and duplicate-candidate count against your existing memories.

preview_blueprint_install

Create or merge user memory from conversation context (journal notes, preferences, plans, reflections, and ongoing personal/project context). This tool can and should be called multiple times in one turn: when a turn surfaces several distinct facts or ideas, record each as its own call rather than combining them into one large memory - one idea per node makes the graph findable later. Not intended for one-off calculations, general knowledge answers, weather, or web lookup results. Writes go to the user's own Empirical memory store. When dueAt is set, the server also sends the user email: a confirmation with a calendar attachment at set time and one reminder when it comes due.

record_graph_memory

Restore a file that was deleted from a memory, if it is still inside the 30-day retention window. Find the id with list_memory_attachments and includeDeleted set to true.

restore_memory_attachment

Search the user's previously stored memories by semantic query and/or filters (category, tags, dates, reminder state, location, graph fields). Reads only the user's own Empirical memory store; not for weather, web lookup, or general knowledge. When working inside a project or repository, pass its name in tags so recall is scoped to that project; if nothing matches, the response includes a retrievalNote and the closest matches from the rest of the workspace, which may belong to other projects. For every record of one kind (all my X), set tags to that topic so the full matching set is returned rather than a ranked subset; list requests return up to 50 records per call, and when hasMore is true call again with the returned cursor until it is false.

query_memories

Set current workspace selection for this OAuth client. Requires a key from list_workspaces, passed as workspaceKey. A missing key is an error, not a reset to the default - to select the default workspace, pass its key explicitly.

set_current_workspace

Give a workspace its kinds of item, their stages and their fields, which is what turns it from a list of memories into a board or a pipeline. ASK THE PERSON WHAT THEY TRACK AND SHOW THEM THE PLAN BEFORE CALLING THIS: a kind of item takes its permanent key from its singular label, and later you can ADD kinds, stages and fields but never remove one. Only for a workspace that has no shape yet. After this, pass itemType when recording memories here, because nothing is classified automatically from the caller's side.

set_workspace_shape

Start a guided interview. Returns the first question and a sessionId. A setup that uses Empirical Tokens first returns a cost preview; tell the person the cost, and call again with acceptEmpiricalTokenCost true only after they agree. The setup_workspace setup, when list_guided_setups shows it, interviews the person to create a shared workspace, or to re-shape an existing one when reshapeWorkspaceId is given.

start_guided_setup

Uninstall a blueprint. Untouched memories are deleted; anything you edited is preserved (detached from the blueprint, not deleted).

uninstall_blueprint

Update an installed blueprint to a newer version when requested by the user. Untouched memories from the old version are removed from active recall along with graph edges to those memories; anything you edited is preserved (detached from the blueprint). This tool has no undo operation for the removed memories or edges.

update_blueprint

Add, remove, or replace relationships on a memory node accessible in the current workspace. Removal discards matching edges; replace=true overwrites the entire relationship list. This tool has no undo operation. Use only for relationship changes requested by the user.

update_memory_relationships

Update an existing stored memory resolved by semantic query or explicit memoryId. A patch can overwrite the saved summary, tags or payload; this tool has no undo operation. Use only for memory edits requested by the user, not external information tasks. It writes only to the user's own Empirical memory store; it does not contact any other system, send notifications, or schedule reminders. If the target memory was created earlier in this same conversation, pass its memoryId instead of query -- search indexing can lag a few seconds behind a fresh write.

update_memory_by_query