AI Directory: find tools by task and evidence

Find documentation profiles, inspect limitations and browse the full catalog. Connect your task to a user-run workflow with explicit checks.

5,246 catalog names, not a count of independently tested products. Some names refer to the same product. Documentation profiles show their own evidence dates; other listings have unknown review status.

Jump to catalog search

Start with a task you can check

Use invented examples first. You run any external AI tool yourself; these workspaces do not call a model. A passing check covers its stated rules, not overall correctness.

Clean a CSV or categorize a catalog

Keep an untouched input. Define allowed categories, preserve exact IDs and prices, and compare the returned rows against those rules. Missing rows or altered prices need review even if the categories look plausible.

Try the synthetic catalogue workflow Open CSV cleanup Read the worked guide

Turn meeting notes into actions

Select the source quote before writing an action. Preserve who said it; leave owners and dates unknown when the notes do not establish them. Review every attribution before sharing the list.

Open the action-item editor Follow the meeting-notes guide

Write from sources and compare outputs

Give both tools the same source and instructions. Record each output’s origin, then judge its claims against the source. Different wording can still be accurate; a matching phrase alone does not prove a claim.

Open the same-task comparison lab Read the verification guide Try a bounded output check

Need a tool with specific constraints? Use the dated, source-based task-to-tool selector. Unknown evidence does not satisfy a requirement.

Documentation-reviewed collection

Find a tool for your next workflow

Choose a task, inspect the limitations, and try a small example before committing. These 12 profiles cite primary documentation reviewed on 15 October 2026. They are not hands-on reviews or rankings.

Browse by task and compare documented profiles

12 matching tools

Build with agents
Goose
Desktop and terminal agent for coding and multi-step workflows.
Build with agents
OpenCode
AI coding agent with terminal, desktop and IDE interfaces.
Work with documents
AnythingLLM
Document chat workspace with local and cloud model connections.
Work with documents
Docling
Document conversion toolkit for structured text, tables and reading order.
Work with documents
Marker
Convert documents into Markdown and JSON with layout and OCR processing.
Work with documents
MinerU
Document parsing toolkit that turns PDFs and office documents into Markdown or JSON.
Work with documents
Kreuzberg
Document extraction library with language bindings, CLI and server interfaces.
Work with documents
Surya
OCR toolkit for text recognition, layout and table analysis.
Evaluate AI
DeepEval
Python evaluation framework for test cases, metrics and agent trajectories.
Evaluate AI
Ragas
Evaluation library for AI applications, including retrieval-augmented generation.
Evaluate AI
Inspect AI
Framework for model evaluations composed from datasets, solvers and scorers.
Evaluate AI
Garak
LLM vulnerability scanner using probes and detectors to surface failure modes.

Browse catalog listings

Search the entire collection by name, task words or category. Multiple words must all match. Results are alphabetical, not quality rankings. Legacy listing claims have not been independently verified.

Search all listingsCategoryDocumentation profiles only

Reset filters

5246 matching listings · Page 27 of 146 · 36 shown

Chat & LLMs

Chatbot Arena

Blind side-by-side model comparisons and leaderboard.

Legacy listing · evidence status unknown