Archive Discord messages. Search them locally. Share a separate, read-only viewer.

Get started · Search only · Your data · Release timeline

Workspace	What you can do
Browse	Search message text, filter by server, channel, author or date, and open saved profiles.
Scrape	Join servers, scan readable channels, pause/resume collection, monitor messages and save profiles.
Stats	Explore activity and contributors, then open the matching messages.
Search only	Run the viewer with an existing archive, without Discord credentials or collection controls.

Get started

Use Python 3.12 or newer with SQLite's FTS5 trigram support. Collection requires a Discord token. Searching an existing archive does not. Read the use warning in the license before collecting data.

git clone https://github.com/k0shir0/searchcord.git
cd searchcord
python -m venv .venv

Activate the environment:

# Linux / macOS
source .venv/bin/activate

:: Windows Command Prompt
.venv\Scripts\activate

Install and start:

python -m pip install -r requirements.txt
python app.py

Open http://127.0.0.1:8000. Windows users can also run start.bat. A first upgrade of a large archive may take several minutes. The browser opens when the database is ready.

Windows launchers discover .venv, then venv, then a working py -3, then python on PATH. They run from their own repository directory, including paths with spaces. Install dependencies into the selected interpreter with run-python.bat -m pip install -r requirements.txt.

Both startup commands show a white terminal banner with large mint Searchcord lettering. Wider windows use a larger seven-row block font with clearer letter shapes. In supported interactive terminals it stays at the top, while startup and runtime logs scroll below it. The banner fits the terminal's starting width and height; narrow or short windows use a smaller layout. Scrolling returns to normal on exit. Redirected output, TERM=dumb and NO_COLOR use a plain ASCII header without terminal controls. Server worker processes keep ordinary log output so they do not redraw or reset their shared console. The collector logs its save policy, request/retry intervals, mandatory breaks and live polling interval once at startup. Saving happens after each fetched page. Scrape requests wait a random 0.5–1 second after the previous scrape response, with a 60-second break after every 100 successful responses.

The collector opens the operating system's default browser after its health check succeeds. No browser executable paths are required. If launching fails, the app keeps running and prints its URL for manual opening.

Use SearchCord

1. Open the settings gear, name your Discord token and select save & use token.
2. Open Scrape, select connect Discord, then choose channels or direct messages.
3. Add conversations to the queue and select start scraping. Leave the message limit blank for full history. Later runs resume or collect newer messages.
4. Open Browse to search. Choose filter suggestions or paste Discord IDs. Date filters use UTC and include the selected end date.

Click a message author to open their saved profile. Fetch expanded profiles saves missing profile details during collection; collect missing profiles and refresh stale profiles update authors already in the archive. Both can be stopped.

Scrape also accepts invite links or codes in a stoppable server-owned queue, with join attempts at least 30 seconds apart. Double-click a server, press Shift+Enter, or choose scan & scrape to queue readable channels and accessible threads automatically. Automatic scans skip names containing bot; those readable channels remain available for manual collection.

Pause scraping saves the current in-flight page before pausing. Resume scraping continues the same job, channel, cursor and remaining message limit. Reloading the page recovers the job. After a server restart, saved scrape and invite queues and live monitors recover paused. Select their original saved account and resume them when ready; no upstream action restarts automatically. Request pacing and mandatory breaks are shared across scrape jobs and cannot be bypassed by pause/resume or starting another job. See collection controls and limits for the full behavior.

Search results load one page at a time. Collection retries temporary failures and keeps saved messages when stopped. Re-fetched IDs update searchable content and preserve observed revisions; confirmed deletions retain an archive tombstone.

Browse uses message-ID cursors to avoid counting every matching message. It shows the current page and whether more results are available, rather than an exact match total. Existing counted page links still work. Reloaded or shared cursor links preserve their page; when earlier cursors are unavailable, first page returns to the beginning. Queries accept up to 200 characters.

Search only

Run the separate frontend in static/search with an existing SearchCord archive:

python search_app.py --db /path/to/archive.db

Open http://127.0.0.1:8001. You can also use --data-dir /path/to/folder when the file is named searchcord.db. Windows users can pass the same arguments to start-search.bat. Without either option, the reader opens data/searchcord.db.

Cursor reads in both frontends use archive_reader.Archive, independently of web setup or collector startup. Standard archives and packed snapshots share the same search/profile read interface and public message-ID cursors.

For a smaller file to deploy, export to a new destination:

python search_snapshot.py data/searchcord.db packed/searchcord.db
python search_app.py --data-dir packed --snapshot

The export preserves message text, names, timestamps, image links and saved profiles. It excludes saved tokens and collection settings. After collecting more data, export a new file and restart the viewer against it. Use the original database for collection.

To prepare a standalone folder containing only the search service and its assets:

python deploy/search/package.py dist/searchcord-search

Copy that folder and your chosen database to the server. Follow the included deployment guide to run the search-only viewer directly with Python or build and run its Docker container. Both options serve existing archives; scraping and archive administration use the Python collector (app.py). The viewer needs a Python server; a static host cannot run it on its own.

Your data

• The collector stores messages, profiles and saved tokens in data/searchcord.db. Saved credentials live in a separate encrypted vault outside the archive directory. Windows protects it with the current user's DPAPI identity; other systems use a separate private encryption key. Keep archives and backups private.
• To change the collector's location, set SEARCHCORD_DATA_DIR before launching. The search service also accepts SEARCHCORD_DB for a specific file. Relative paths resolve against the application folder.
• Both apps start on loopback and have no built-in login. Put access control and HTTPS in front of a hosted viewer. Share only data you are entitled to share.
• New collection retains full received message JSON and attachment metadata. The collector also saves bounded local copies of attachments, avatars and banners beside the database in media/. Missing or oversized downloads are recorded explicitly. Export with --with-media to preserve available binaries in the viewer; refresh inaccessible remote links through the collector.
• An older archive upgrade keeps a recovery backup. Allow extra disk space and check the upgraded archive before removing a backup.
• Clear All Data removes archived messages and derived metadata after confirmation. Stop active and paused scrapes, live monitors and DM cleanup first. Profile jobs stop during clearing; saved credentials remain.
• remove all saved tokens removes the sealed credential vault and saved credential settings after a second-click confirmation. Stop scrapes, DM cleanup and live monitors first; invite joining and profile backfill stop during removal. Messages and unrelated settings stay. Save a new token to connect again. Closing Settings cancels confirmation.

Old databases, recovery backups and SQLite free/WAL pages may still contain credentials saved before this upgrade. They are not securely erased by migration or token removal. Keep those copies private and rotate retired credentials.

Recovery backups are named searchcord.db.pre-compact-*.bak.gz (or .bak if compression fails). Newly generated recovery copies remove credential settings and retain private messages; older copies may contain plaintext tokens. Migration checks SQLite and full-text integrity. Fields already discarded by an older migration can only be recovered from an original archive or backup. Test upgrades on a separate copy first.

Documentation

• Collection, invites, scanning and pause/resume
• Saved profiles and backfill
• Read-only search behavior
• Standalone deployment
• Archive reliability, credentials and remaining limits

The production branch contains the application, required assets, deployment files, licenses and operating documentation. .gitignore explicitly lists the allowed files. Keep local checks, benchmarks, review reports, experiments and generated output ignored; add new production files to the allowlist before committing them.

Release timeline

Major development milestones in the current main history. Dates use America/Chicago and reflect commits, not deployment dates. As of 2026-10-09, no versioned GitHub Releases have been published.

Date	Update
2026-10-09	Durable paused job recovery, observed message revisions, local attachments and protected saved credentials.
2026-10-09	Invite queues, readable server/thread scans and reload-recoverable scrape pause/resume, with enforced request pacing and breaks.
2026-10-07	Portable Windows startup, terminal banner and saved-token removal.
2026-10-02	Faster cursor-based search and standalone viewer packaging with Docker support.
2026-09-30	Compressed search archives that exclude saved tokens.
2026-09-29	Immediate profile collection while scraping and stoppable, retryable profile backfill.
2026-09-28	Separate read-only viewer, archived profile cards and more reliable scraping.
2026-09-26	Multiple saved accounts, confirmed archive cleanup and expanded server/contributor rankings.
2026-09-25	Reduced archive storage and compressed recovery backups.
2026-09-24	Unified scraping tools and activity charts.
2026-09-23	New search layout, filters, resumable scraping and DM message cleanup.
2026-07-23	Unified web app with live monitoring and ChatML export.
2025-09-20	Initial scraper and viewer.

License

MIT, with a wrongful use warning.