Skip to content

Auto Archiver (Bellingcat)

FreeFirst-line

Publisher: github.com/bellingcat/auto-archiver | Type: Non-detector | Cost: Free

TL;DR

Bellingcat's open-source pipeline for automatically preserving social-media URLs as evidence. URLs go in (from a Google Sheet, a CSV, or the command line); archived files plus metadata hashes come out (to S3, Internet Archive, Google Drive, or local storage). The infrastructure layer underneath every piece of source-history work where evidence preservation matters.

What it does

Auto Archiver is a Python pipeline that takes URLs from a source (typically a Google Sheet, but also CSV files or direct command- line input), archives the content of each URL, and writes the archived artefact plus a metadata record to a configured back-end (S3 bucket, Google Drive, Internet Archive, or local storage). The tool handles social-media posts, videos, images, and webpages, and produces both the archived file and a verifiable hash record that supports chain-of-custody work.

A status report can be appended back to the original Google Sheet, which makes the tool natural to deploy as the archiving layer underneath a fact-check workflow that uses Sheets for intake. The pipeline is published on PyPI and as a Docker image with monthly maintenance; v1.2.6 shipped 7 April 2026, the project carries 44 releases and active contributor activity.

For toolkit work, Auto Archiver is the rapid-archiving WACZ- adjacent workflow that the toolkit identified as a high-priority gap. A Pillar 2 fact-check operation that processes user-submitted links can run those links through Auto Archiver on intake; by the time the editorial team reaches a verdict, the source content has been preserved in a form that survives platform takedown, account deletion, and post-publication editing.

When to use it

  • A regional fact-check operation needs an automated archiving layer that runs on Google-Sheets intake and writes to S3 or Internet Archive without manual per-URL work.
  • An institutional investigation is processing dozens or hundreds of social-media URLs and needs each one preserved as evidence before any analysis begins.
  • A verification workflow is preparing for defamation defence or platform appeal and the source URLs must be archived with verifiable metadata hashes before any publication.
  • A research project is collecting longitudinal evidence (an ongoing CIB campaign, a recurring deepfake actor, an election- period narrative cluster) and needs continuous archiving over weeks or months.

Limitations

Limitations

  • Requires a Google Cloud Service Account plus JSON credentials for the Sheets-driven workflow. Setup carries an institutional-IT cost that frontline fact-checkers without admin access cannot complete on their own.
  • Frequent platform-blocking adaptation needed. The pipeline relies on platform APIs and scrapers; when Twitter/X changes its rate limits or Facebook tightens scraping protections, the maintainer ships a fix, but a self-hosted deployment can lag the upstream by weeks if the operator does not redeploy promptly.
  • Docker or Python install path; the tool is not a one-click product. Operators need command-line comfort and a server or workstation to run the pipeline.
  • The archived artefact's evidentiary status depends on the back-end's chain-of-custody discipline. Uploading to a public S3 bucket without access controls is a different evidentiary posture from uploading to a controlled S3 bucket with documented access logs; the tool ships the artefact, the operator owns the chain.

Privacy and threat model

Auto Archiver is self-hosted. The pipeline runs where the operator deploys it; the source URLs and archived artefacts pass through whatever infrastructure the operator chooses. This is the strongest privacy posture in the 2A.2 cell because the operator controls the data flow end to end; Bellingcat operates the upstream code repository but does not host any archive on behalf of users.

The privacy considerations move to the chosen back-end. An S3 bucket on AWS sits under US/Amazon jurisdiction; an Internet Archive submission is public and persistent; Google Drive is under Google jurisdiction with the operator's account permissions; local storage stays on the operator's machine. For Sri Lanka and Laos surveillance-environment work, the operator chooses the back-end to match the threat model: local or trusted-jurisdiction storage for source-identifying material, public Internet Archive for material where preservation is the goal regardless of who reads it.

Source-protection override (S3 — graphic, sexual, child-safety, or abuse material)

Auto Archiver pipelines that ingest conflict footage, alleged abuse imagery, or non-consensual intimate content place archived copies on whichever back-end is configured — and any future reviewer of the archive is exposed to the same material. Mitigation steps:

  1. Before archiving any conflict, violence, or abuse content, follow newsroom protocol on whether the material may be preserved at all; some categories (child sexual abuse material) carry legal obligations that override editorial preservation interest.
  2. Minimise reviewer exposure: rotate analysts, document sensitivity classifications in archive metadata, and route the archive away from team-wide shared storage by default.
  3. Where preservation is required for legal or evidentiary purposes, prefer back-ends with documented access controls (encrypted local storage, restricted S3 bucket) over public Internet Archive submission.
  4. Apply the staff-safety protocol on repeated viewing for any analyst working through a sensitive archive batch.

Country and platform applicability

  • Indonesia: language-agnostic; usable as the archiving layer underneath the MAFINDO/CekFakta workflow at Kalimasada. No documented Indonesian-specific deployment, but the tool is operationally portable.
  • Laos: language-agnostic; for regional partners archiving Lao-content links, the self-hosted deployment is the right architectural fit because the operator controls whether any data crosses Lao state-aligned infrastructure.
  • Malaysia: language-agnostic; useful for Sebenarnya.my and partner deployments that need automated archiving of user-submitted links.
  • Philippines: language-agnostic; pairs with #FactsFirstPH coalition workflow for evidence preservation across coalition members.
  • Sri Lanka: language-agnostic; useful for Watchdog and Hashtag Generation election-period archiving. Self-host on local infrastructure for surveillance-aware posture.
  • Thailand: language-agnostic; pairs with Thai PBS, SONP, and Cofact archiving needs.

Platform applicability: Facebook, Facebook Groups, TikTok, YouTube, Telegram, X. Twitter/X support requires the v1.2.3+ drop-in alternative; LINE-shared content is platform-restricted and not directly archivable through Auto Archiver.

How to access

Open-source under MIT licence at github.com/bellingcat/auto- archiver. Install via Docker (docker pull bellingcat/auto-archiver) or pip (pip install auto-archiver). Documentation lives at auto-archiver.bellingcat.com. The Bellingcat introductory article on the tool is published at bellingcat.com. Latest release v1.2.6 (7 April 2026) per the project's GitHub releases page; the maintainer ships roughly monthly with bugfix and platform-adaptation releases.

Cost (current as of 2026-05)

Free under MIT licence. The structural cost is operational — infrastructure to run the pipeline (a Docker host or Python environment), the back-end storage (S3, Google Drive, Internet Archive submission, or local), the institutional setup of a Google Cloud Service Account if Sheets-driven intake is used, and the operator time to redeploy when upstream releases ship platform-adaptation fixes.

Quickstart

  1. Decide the deployment posture: Docker (docker pull bellingcat/auto-archiver) for production, pip install for local development, manual command-line runs for ad-hoc work.
  2. For Sheets-driven intake, create a Google Cloud Service Account and download the JSON credentials; share the target Google Sheet with the service account's email.
  3. Configure the orchestration YAML (back-end storage, Sheet ID, archiving rules) per the Bellingcat documentation.
  4. Run the pipeline: docker run -it --rm -v secrets:/app/secrets bellingcat/auto-archiver --config secrets/orchestration.yaml.
  5. Confirm archived artefacts and metadata hashes appear in the chosen back-end.
  6. Schedule the pipeline (cron, systemd, or a CI scheduler) for continuous archiving on intake.
  7. Watch the Bellingcat releases page; redeploy when a release ships platform-adaptation fixes (Twitter/X, Facebook, TikTok adaptation has been the recurring maintenance pattern).

In the toolkit's workflow

Source-history pillar non-detector infrastructure layer per Architectural Anchor 1. Sits in 2A.2 nominally as an "enhanced reverse-image / geolocation" cell entry but functionally as the evidentiary preservation layer that runs underneath the cell's visual-verification work. The cell groups Auto Archiver here because every reverse-image hit, every GeoSpy hypothesis, every YouTube Data Viewer extraction is more defensible when the source URL has been preserved before analysis began. Cross-cell awareness with the Digital Safety section – Auto Archiver's archiving-as-evidence pattern is foundational there.

Standard combinations:

  • With InVID-WeVerify at 2A.2 / 1B.1 – InVID's WACZ module captures a single URL on demand; Auto Archiver preserves bulk URLs from Sheets as a continuous workflow.
  • With GeoSpy at 2A.2 – Auto Archiver preserves the source URL before GeoSpy's geolocation hypothesis is generated, so the eventual fact-check has the original context in evidence.
  • With Citizen Evidence Lab YouTube Data Viewer at 2A.2 when YouTube content needs both upload-time verification and bulk archiving for a longitudinal record.
  • With Meedan Check at 2B.1 when tipline intake generates a continuous stream of URLs that need preservation before the editorial workflow processes them.
  • With CIB Mango Tree at 1C.1 when behaviour- pillar analysis needs preserved historical artefacts to support the network finding.

Decision-tree references: T2 Video triage and T6 Source-protection both route through Auto Archiver as the preservation step before any analysis. The Digital Safety section references Auto Archiver as the evidentiary baseline alongside Browsertrix (the WACZ-native alternative kept at the Digital Safety reference layer rather than promoted into 2A.2; see change-log).

This card carries no detector signal class: Auto Archiver preserves source content as evidence; it does not produce a content-classification verdict. Per Anchor 1 the preserved artefact is the substrate for downstream source-history analysis; per Anchor 2 archiving by itself is not a publishable claim, but its absence weakens any claim that depends on content the platform later removed.

Override notes

Override notes

  • Rapid archiving WACZ workflow (rapid-archiving-WACZ- workflow-pointer): the tool implements the rapid-archiving WACZ-adjacent workflow it recommends. Operators preserve source content as evidence on intake; the toolkit treats archive-on-intake as the recommended pattern for any Pillar 2 operation that may need defamation defence, platform appeal, or longitudinal-cluster analysis.

Sources