kristenpan/pickoo-your-favorites

Current version: 1.0.1

**Throw all the fun in. **

An AI bookmark that lets local agents automatically organize your content and preserve your inspiration.

Pickoo is a local-first AI favorites and inspiration organizer workbench. The interface uses a desktop browsing style of “card stack → grouped → single read per item,” with editable Markdown and evidence packs at the data layer, not tied to the content platform. It is suitable for content creators and anyone who is used to getting inspiration through collection, organization, and review.

Product Features

Local First, Native Agent

Original texts, media, transcripts, and breakdown results are stored in users’ own local databases, using editable markdown and evidence packs as long-term entry points. Codex, WorkBuddy, Hermes, Claude, and Claudian share the same data and task status, allowing users to continue retrieval, organize, review, analyze, or create directly without repeatedly uploading materials between different AI tools.

Pickoo prioritizes users to reuse their existing agent, local OCR, and ASR capabilities, without forcing binding to Pickoo’s own cloud AI services. Actual costs still depend on the agent, model, and local runtime environment chosen by the user; “Local priority” does not mean completely offline when scraping public links.

Saving Actions Makes Things Easy to Use

Paste links in Pickoo, or directly hand over publicly accessible links and platform shared text to already integrated agents. Through messaging channels like Hermes, content can also be submitted from WeChat or Feishu; Different entry points ultimately lead to the same task center and local database.

Pickoo is responsible for source identification, plagiarism checking, scraping, and subsequent processing. Platform login, membership, DRM, region, and access restrictions remain in effect; Content that cannot be fully obtained will be clearly marked as pending review, partially completed, or failed.

Break down the content clearly, not just summarize

Videos and podcasts can generate timestamped transcripts and semantic chapters, and images can perform OCR. Each source can further refine six categories of reusable content assets; Whether used to organize interests, study materials, or creative inspiration, only candidates that have passed evidence review and meet quality thresholds are included in the asset library.

Optional speaker separation can distinguish who is speaking during the interview and at when, and supports manual renaming versus cross-episode memory. It does not equate to automatically confirming true identity without basis, and the results are still subject to quality gates.

Preserve original evidence for easier access

Pickoo does not use a single AI summary to replace the original content. Original text, images, audio and video, author and source, and subsequent OCR, ASR, chapters, and asset breakdown are stored in layers; The results of the breakdown can be returned to the corresponding original text, image, time period, or footage. Therefore, the Agent can both call the conclusion and verify where it comes from.

Collecting can also be an exploration game

Users can tag likes, jot down ideas, or save individual images, video frames, and footage clips, adding tags, annotations, and personal preferences. The achievement system unlocks badges based on collection, organization, interaction, and exploration behaviors, gradually turning the library into a personal interest map rather than just another collection list waiting to be cleaned.