Glaze tests from a week of coastal fog. Keeping the failures close.
Your Instagram history.
Actually yours.
Import your archive into private local SQLite. Ask for images in natural language, search OCR and visual concepts, organize Saved posts automatically, and build exportable moodboards.
Home
A visual system built from tide charts, shipping labels, and one stubborn shade of red.
Last light moving across the south wall. 22 seconds, no soundtrack.
Try the full visual-memory workflow.
Explore a privacy-safe sample archive—analyze media, ask in natural language, organize Saved posts, and compose a board. The demo is interactive, but nothing leaves your browser.
- 01AnalyzeOCR, objects, colors
- 02AskHybrid visual search
- 03LibraryClusters, tags, review
- 04BoardsArrange, note, export
Ask your visual memory
One archive in. A whole memory system out.
Request a JSON export from Instagram Accounts Center, point Gramclaw at the ZIP, and open the local workspace. Re-import later—merges are idempotent and preserve newer local data. Every person imports their own export; no maintainer account or shared login is bundled.
How to export from Instagram# Install the downloaded release
$ npm install -g ./gramclaw-1.1.0.tgz
# Create your private local workspace
$ gramclaw init
# Build your local memory
$ gramclaw import archive ~/Downloads/instagram-export.zip
✓ 4,218 posts · 12,904 DMs · 1,103 saved
# Build the private visual index
$ gramclaw analyze run --provider local
✓ OCR · objects · colors · embeddings
$ gramclaw serve --open
→ http://127.0.0.1:4667Instagram is the source. Your machine is the home.
Archive and live transports converge on one normalized local core, so the interface, CLI, backups, and agents all see the same truth.
Private media understanding
Local OCR, image descriptions, dominant colors, objects, visual style, embeddings, and duplicate fingerprints—with progress and retry support.
Ask what you remember
Hybrid caption, OCR, metadata, and semantic-image search with creator, date, type, Saved/Liked, color, and topic filters.
Wood · warm interior · last year
SavedSaved, automatically organized
Topic clusters and a review queue sit beside your own collections, tags, and bulk organization tools.
InteriorsWood, kitchens, quiet rooms
CeramicsGlaze, clay, studio tests
UnorganizedReady for your review
Build a board while the idea is alive
Pull references directly from search, rearrange them freely, add working notes, and export the finished moodboard as an image or PDF.
Backups you can inspect
Deterministic JSONL shards stay Git-friendly. Cookies, tokens, WAL files, FTS shadow tables, and downloaded media stay out.
Ephemeral, preserved
Stories and archive media are copied into your local originals cache, with live downloads paced separately.
Human-friendly in the browser. Agent-native in the terminal.
Every read command emits stable JSON. Your scripts and agents can search, triage, inspect people, read conversations, and analyze the relationship graph without touching Instagram again.
$ gramclaw ask "wooden kitchens I saved last year" --json
{
"items": [{
"id": "post_8f21…",
"score": 86,
"why": ["Objects: wood", "Saved item"],
"analysis": {"colors":["#9a7557"]}
}]
}Your archive never needs a cloud.
The durable path is deliberately boring: JSON export → local SQLite → loopback web app. Live connections are optional adapters, never the canonical store.
Private web endpoints can change or rate-limit automation. Gramclaw surfaces transport health without making live access a dependency.
Good questions.
Plain answers.
Gramclaw is independent software, not affiliated with Instagram or Meta. Use live access gently and only on accounts you control.
Does Gramclaw upload my archive?+
No. The app imports into SQLite and analyzes media on your machine. The public site never receives your archive, images, messages, cookies, or tokens. Optional cloud vision runs only when you explicitly choose it.
Does the download include somebody else’s Instagram account?+
No. The package contains no maintainer username, profile, cookies, messages, Saved items, Liked items, or archive. The public demo is fictional, and every installation creates a separate local workspace for that user.
Do I need an Instagram API key?+
No for archive import, search, DMs, insights, media preservation, backups, and the local web workspace. Optional live reads can reuse your signed-in browser session. Professional accounts can also connect the official Instagram Graph API.
What can it import?+
JSON exports containing posts, carousels, reels, stories, comments, likes, saved posts, followers, following, profile information, and direct-message history. Imports are idempotent and merge-safe.
Can agents use it?+
Yes. Every read path has stable JSON output: visual search, media analysis, Saved collections, boards, conversations, profiles, relationship graphs, and insights.
Can it publish?+
Yes, with guardrails. Browser-cookie transport supports a local JPEG post or story. The official Graph transport supports URL-based posts, carousels, reels, stories, comments, replies, and messages. Writes remain drafts unless you explicitly pass --yes.
Local-first Instagram memory