Every tool

🤖 AI Drafted (Not reviewed)

A tool is one step: it reads something, does one job, and emits a result. Workflows are tools wired together. Each page below says what the tool reads and emits, every option it takes, and — for tools that ask a model — the exact words it sends. There are 126 of them.

Agent

Tool	What it does	Uses a model
Agent Coordinator	Runs multiple agents in parallel and combines their results	yes
CLI Agent	Run a local agent CLI (Claude/Codex) and capture stdout/stderr.	no
ReAct Agent	AI agent that uses tools to accomplish tasks	yes
Supervisor Agent	Orchestrates multiple worker agents to accomplish complex tasks	yes
Swarm Agent	Agent that dynamically routes to other agents based on task requirements	yes

Audio

Tool	What it does	Uses a model
Transcribe Audio	Transcribe audio files to text using Whisper, Apple Speech, or cloud providers	yes

Llm

Tool	What it does	Uses a model
Archival Summary	Compose nine-section archival summary for a folder (leaf node of the Catalogue workflow)	yes
Classify Text	Classify text into categories	yes
Clean Dates (folder)	Pick global canonical dates across all pages in the folder and save a cleaned artifact on the folder doc.	yes
Clean Dates (page)	Group near-duplicate dates within each page using a focused LLM call. Re-points claims at the merged canonical entity.	yes
Clean Events (folder)	Pick global canonical events across all pages in the folder and save a cleaned artifact on the folder doc.	yes
Clean Events (page)	Group near-duplicate events within each page using a focused LLM call. Re-points claims at the merged canonical entity.	yes
Clean Keywords (folder)	Pick global canonical keywords across all pages in the folder and save a cleaned artifact on the folder doc.	yes
Clean Keywords (page)	Group near-duplicate keywords within each page using a focused LLM call. Re-points claims at the merged canonical entity.	yes
Clean Organizations (folder)	Pick global canonical organizations across all pages in the folder and save a cleaned artifact on the folder doc.	yes
Clean Organizations (page)	Group near-duplicate organizations within each page using a focused LLM call. Re-points claims at the merged canonical entity.	yes
Clean People (folder)	Pick global canonical people across all pages in the folder and save a cleaned artifact on the folder doc.	yes
Clean People (page)	Group near-duplicate people within each page using a focused LLM call. Re-points claims at the merged canonical entity.	yes
Clean Places (folder)	Pick global canonical places across all pages in the folder and save a cleaned artifact on the folder doc.	yes
Clean Places (page)	Group near-duplicate places within each page using a focused LLM call. Re-points claims at the merged canonical entity.	yes
Clean Up Text	Clean extracted or transcribed text: fix OCR noise, normalize whitespace and hyphenation, strip page headers/footers and scanning artefacts — preserving the original meaning.	yes
Detect AI Text	Estimate likelihood that text is AI-generated	no
Extract All Entities	Single-pass extraction of people, places, organisations, dates, events, and keywords. One LLM call per page returns all six types as JSON — 6× fewer calls than the per-type extractors, same downstream shape (KG claims + per-page artifacts).	yes
Extract Book Index Topics	Parse a back-of-book index into topic entities and grounded statements from the referenced pages.	yes
Extract Citation Usage	Extract citation usage section only.	yes
Extract Citations	Extract bibliography entries and inline citation links	yes
Extract Dates	Extract dates section only.	yes
Extract Entities	Extract named entities (people, places, dates, organizations) from text	yes
Extract Entities	Extract and persist entity rows only from existing transcription artifacts	yes
Extract Events	Extract events section only.	yes
Extract Geo	Extract place names from text and geocode them to lat/lon	yes
Extract Interpretations	Extract interpretations section only.	yes
Extract Keywords	Extract keywords section only.	yes
Extract Legal References	Extract legal references section only.	yes
Extract Mines	Extract mines section only.	yes
Extract Organizations	Extract organizations section only.	yes
Extract People	Extract people section only.	yes
Extract Places	Extract places section only.	yes
Extract Properties	Extract properties section only.	yes
Extract Quotes	Extract quotes section only.	yes
Extract Rivers	Extract rivers section only.	yes
Extract SVO Claims	Extract and persist SVO KnowledgeClaim rows only from existing entities and transcription artifacts	yes
Interpret	Apply an interpretive framework to text and persist the resulting interpretations (hermeneutic layer)	yes
Key People	Identify important people	yes
Keywords	Extract keywords from text	yes
Language Identification	Detect language(s) in a document and output aggregate metadata	no
Model Comparison	Compare responses from multiple LLM models	yes
Named Entity Recognition	Extract named entities with a selectable backend provider	yes
Questions	Generate questions from text	yes
Rewrite	Rewrite text in different styles	yes
Sentiment	Analyze text sentiment	yes
Split Diary Entries	Split each page’s transcript into per-day entry nodes with a date attribute and the day’s bounding box — the timeline/calendar feed	yes
Summarize	Generate text summaries	yes
Summarize Collection	Generate a high-level summary of an entire collection	yes
Summarize File	Generate a summary of a single document’s content	yes
Summarize Folder	Generate a summary of all documents in a folder	yes
Timeline	Create chronological timeline	yes
Translate	Translate text from source_lang to target_lang.	yes
Translate Review	Second-pass AI double-check of a draft translation against the original source: corrects mistranslations, omissions, and altered names. Target language, model, and prompt are editable.	yes
Translate Text	Translate extracted or transcribed text into the user’s preferred language. Target language, source language, and model are all editable on the node.	yes
Write KG	Persist KG rows from an explicit upstream write bundle	no

Output

Tool	What it does	Uses a model
Export Documents	Export selected documents as Markdown, Word, and/or Excel files.	no
Write File	Save upstream text to a file on disk (per-file or aggregate).	no

Research

Tool	What it does	Uses a model
Browser Navigate	Navigate to a URL and extract page content, title, and links. Sandboxed — only http/https URLs allowed.	no
Document Fetch	Fetch a document from a URL and optionally save as a Layer 1 Source in the Fichero database. Sandboxed — only http/https URLs allowed.	no
Web Search	Search the web using DuckDuckGo HTML (sandboxed). Returns title, URL, snippet, and relevance for each result.	no

Source

Tool	What it does	Uses a model
Annotations Source	Iterate a document’s user highlights / notes / regions and emit one cropped input per annotation. Use upstream of any vision or LLM tool to make the AI operate on a marked region instead of the whole document. (#914)	no
Artifacts Source	Emit a document’s existing artifacts as text — a transcription, a review, a translation — so a downstream tool runs on what an earlier step produced instead of on the page image. Pick the artifact type; the newest of that type wins unless you ask for all.	no
Book Structure	Extract chapter/section/subsection ranges from a book PDF outline	no
Collection	Get all files from a library collection	no
Detect Structure	Detect chapters, sections, and subsections for a book PDF	no
Files	Pass through input files from workflow context	no
Folder	Get files from a specific folder	no
Search	Find files matching a search query	no
Selection	Use the currently selected documents from the library UI	no
Split Chapters	Split a book PDF into chapter documents using TOC and heading cues	no

Text

Tool	What it does	Uses a model
Merge Geometry	Place a reviewed transcription’s words on the measured word boxes from an OCR pass, so corrected text becomes clickable on the page. Every word records whether its box was measured or interpolated, and a page whose line structure cannot be trusted is refused rather than given a confident wrong overlay.	no

Transform

Tool	What it does	Uses a model
Adaptive Binarize Images	Locally clean uneven scan backgrounds into black and white.	no
Aggregate	Combine upstream fan-out results into a single payload.	no
Auto Crop Scan Borders	Detect and crop dark scan margins locally.	no
Auto Deskew Images	Detect and straighten skewed scans locally.	no
Consistency Check	Flag deterministic numeral, name, and formula inconsistencies.	no
Denoise Images	Locally denoise selected scans without modifying sources.	no
Enhance Images	Create contrast/sharpness/denoise image derivatives without modifying source files.	no
Extract Date	Extract the historical date a document was written (Gregorian, Julian, French Republican, Old Style, regnal, era names) and store it as a Julian Day Number range for sorting and filtering.	no
Fuzzy Clean Images	Despeckle and clean scan background without modifying source files.	no
OCR Cleanup	Fix common OCR artefacts: dehyphenate line-break splits, rejoin multi-column text, strip library ownership stamps.	no
Prepare Images	Normalize images/PDF pages for OCR without modifying source files.	no
Recombine Segments	Stitch segment image files into one derived image.	no
Remove Background Images	Create alpha-background image derivatives without modifying source files.	no
Rotate / Auto-Orient Images	Create rotated or EXIF-oriented image derivatives without modifying source files.	no
Segment Images	Detect foreground document regions and emit cropped segment derivatives.	no
Split Images	Split images into grid tiles or PDFs into page images without modifying sources.	no
Text Reflow	Join soft-wrapped lines within paragraphs, preserve paragraph structure, and de-hyphenate words split across line breaks. Optional AI refinement pass available for edge-case word joins.	yes
Zoom	Crop and magnify image regions or line strips.	no

Utility

Tool	What it does	Uses a model
Import Artifacts	Register import-receipt and transcription artifacts only	no
KG Persist / Finalize	Finalize corroboration counts, canonical KG embeddings, and graph snapshot from existing KG rows	no
Merge / Dedup	Apply existing entity-resolution rules, claim suppression, and trivial-claim pruning to existing KG rows	no
Organize Same Documents	Create one subfolder per duplicate cluster and move members into it via the audited action registry	no

Video

Tool	What it does	Uses a model
Describe Video	Describe video content visually and transcribe audio track	yes

Vision

Tool	What it does	Uses a model
AI Convert	Generative format conversion: an AI vision model re-creates the page as Markdown, HTML, SVG, LaTeX, or CSV. The output is the model’s reading of the page, not a mechanical file conversion.	yes
Analyze	Custom vision analysis	yes
Caption	Generate short image caption	yes
Classify	Categorize document type	yes
Classify Script Type	Detect whether a document is typescript, manuscript, HTR, or paleography	yes
Colors	Extract color palette	yes
Compare	Compare multiple images	yes
Describe	Generate image descriptions	yes
Detect Regions	Finds WHERE the words are, on-device and free. Apple Vision reads the page locally to locate line and word boxes, so it does produce text — that text is a by-product of finding the boxes, not a transcription: it never replaces the page’s transcript, and no model is called. Runs before a transcriber so every box exists up front.	no
Diagram	Parse diagrams and flowcharts	yes
Economy HTR	Free Apple Vision line boxes + a cheap local HTR backend (no paid API calls).	no
Extract	Extract multiple fields in one call	yes
Faces	Detect and describe people	yes
Handwriting	Historical handwriting transcription	yes
Layout	Analyze document structure	yes
Objects	Detect and identify objects	yes
Quality	Assess image quality	yes
Safety	Content safety assessment	yes
Scene	Classify scene environment	yes
Similarity	Score image similarity	yes
Style	Classify artistic/document style	yes
Table	Extract tables from images	yes
Tags	Generate keyword tags	yes
Transcribe	Extract text from images (OCR)	yes
Transcribe Review	Second-pass QA of a prior transcription against the image	yes

Workflow

Tool	What it does	Uses a model
Sub-Workflow	Run a child workflow behind declared input/output contracts.	no