Weave

Permanently delete a memory from the user's vault. Use Search & Browse Memories to find the right `submission_id` first. This is destructive: it removes the memory and the facts only it supported; facts also known from other memories stay, and a shared summary may still paraphrase it for a while. This tool cannot bring a forgotten memory back — the user confirms any restoration explicitly in the Weave app.

weave_forget_memory

Update the content, primary tag, or additional tags of an existing memory. Look up the memory's `submission_id` with Search & Browse Memories first. Pass only the fields you want to change. Passing `additional_tags: []` clears all additional tags. Visibility cannot be changed after creation. Content is capped at 10,000 characters. Saving is instant; memory processing continues in the background, so a just-edited memory may be returned as pending and search may briefly show the older wording — don't retry an edit because search doesn't show it yet.

weave_edit_memory

Return all memories carrying a specific additional tag, optionally narrowed by primary tag. Use View All Tags first if you're unsure which tags exist. Each item carries `ingestion_status`: queued/processing (still being processed), indexed, or failed (see `ingestion_error`). Saving is instant; memory processing continues in the background, so very recent saves may be returned as pending — don't retry a save because search doesn't show it yet. Pagination: **on the first call, omit `cursor`.** Each response includes up to `limit` items (default 50, max 200), the `total_count` of matches, and a `next_cursor` token. To fetch the next page, call again with the same `additional_tag` plus `cursor` set to the previous response's `next_cursor`. When `next_cursor` is null, you've reached the end.

weave_get_by_tag

Rename an additional tag across every memory that uses it. Look up the tag's `tag_id` with View All Tags. The new name must not collide with an existing tag (case-insensitive) and cannot be one of the reserved primary tags (YOU, WORK, PEOPLE, LIFE).

weave_rename_tag

Save a memory to the user's personal Weave vault so they can recall it later across conversations. Use this for preferences ("I prefer aisle seats"), facts ("my preferred airport is SFO"), goals, meeting notes, ideas — anything the user wants persisted. Pick a `primary_tag` from YOU (about the user), WORK, PEOPLE (about others), or LIFE (everything else). Add `additional_tags` for free-text labels like 'travel' or 'recipes'. Set `agent` to your own name (e.g. 'claude') so the user can see which agent created which memory. Content is capped at 10,000 characters — split longer notes into separate memories. Saving is instant; memory processing continues in the background, so very recent saves may be returned as pending — don't retry a save because search doesn't show it yet. If Weave refuses the save because the user previously asked it to forget this memory or the person it names, do not retry or rephrase — tell the user, and let them confirm restoration in the Weave app.

weave_create_memory

Browse the user's saved memories with optional filters. Pass `q` for a full-text search; combine with `primary_tag`, `additional_tag`, or date ranges (`since`/`until`) to narrow further. Each item carries `ingestion_status`: queued/processing (still being processed), indexed, or failed (see `ingestion_error`). Saving is instant; memory processing continues in the background, so very recent saves may be returned as pending — don't retry a save because search doesn't show it yet. Pagination: **on the first call, omit `cursor`.** Each response includes up to `limit` items (default 50, max 200), the `total_count` of matches, and a `next_cursor` token. To fetch the next page, call this tool again with the same filters plus `cursor` set to the previous response's `next_cursor`. When `next_cursor` is null, you've reached the end.

weave_list_memories

Find memories by meaning, not just keyword. When `q` is provided, runs meaning-based search over the user's processed memories and returns the best matches (each is a whole saved memory) plus, by default, a short context block assembled from the user's whole memory graph. Use this for natural-language questions ("what diet does the user follow?"). **Calling with no `q` returns the most recent memories** — useful as a first look at the user's vault before you know what to search for. Results are not paginated: `next_cursor` is always null; raise `limit` (max 50) or narrow with `primary_tag` instead. Saving is instant; memory processing continues in the background, so very recent saves may be returned as pending — don't retry a save because search doesn't show it yet. Very recent saves may be returned flagged `pending` (matched by text, not yet by meaning), and `pending_count` says how many of the user's memories are still processing; a recently changed fact may still show its older version until processing catches up. For exact-text / tag / date filtering, prefer Search & Browse Memories.

weave_search_memories

Send one plain-text SMS to the user's OWN phone — the one linked to their Weave account via Needle. Use this to deliver a result the user asked to receive as a text (a daily briefing, a reminder that a task finished). There is NO recipient parameter: this tool can only text the account owner, never a third party. The message is delivered as written (plain text, no markdown; leading/trailing whitespace is trimmed) and queued durably, arriving within moments. Limited to 10 texts per hour. If the account has no phone linked, the call fails with linking instructions — relay them to the user.

weave_send_message

List every additional tag the user has created, with the number of memories under each. Useful before filtering with Filter Memories by Tag, or for showing the user an overview of how their vault is organized.

weave_list_tags