BagIt Integrity Report

Get one structured integrity report for a bounded BagIt ZIP without installing a validator in your calling application. This experimental Actor checks the complete supported payload inventory and every available listed payload/tag target against its checksums. It reports missing, unexpected and changed files with coverage and acquisition provenance.

Use from an AI agent through MCP

Add this URL to a client that supports remote HTTP MCP, then authorize with your own Apify account using the Apify MCP setup guide:

https://mcp.apify.com/?tools=l3digital/bagit-integrity-report

This configuration selects the Actor directly. Availability still depends on Apify account and Actor eligibility. Call l3digital/bagit-integrity-report with the example input below; the pricing and input restrictions on this page apply.

When using call-actor, its response contains run status and storage IDs. If the run is still active, check that run with get-actor-run. After success, retrieve the report with get-dataset-items using the returned dataset ID (defaultDatasetId in the run API). These retrieval tools load with the Actor. Retrieve the existing result instead of starting another run. Inspect the report status and the coverage or refusal fields described below before using its values.

Try the example

The default input runs a tiny built-in public-domain synthetic bag through the same complete validator:

{"mode": "demo"}

The example performs no network request, returns example: true and useful: false, and never charges a report event. It demonstrates correctness; it is not external demand or a repository acceptance decision.

For an authorized public ZIP, use:

{"mode": "url", "url": "https://example.org/your-bag.zip"}

Replace the placeholder with your own authorized public package URL. Only HTTPS is accepted. Queries, fragments, userinfo, explicit ports and non-public DNS answers are refused. Redirects receive the same checks. No source credentials, cookies, proxy configuration or caller headers are forwarded.

The Console form lists both fields. The Actor validates their relationship at runtime: url is required in URL mode and must be omitted entirely in demo mode, including any explicit null value. Invalid input fails before acquisition or report charging. The platform form schema does not express these conditional rules.

Supported packages

• BagIt 1.0 with UTF-8 tag files, packaged at ZIP root or under one enclosing directory.
• Stored or deflated ZIP entries, at least one SHA-256 or SHA-512 payload manifest, and a data/ directory. Both allowed algorithms and corresponding tag manifests are checked when present.
• Optional bag-info and ordinary tag files within the fixed limits.

Other versions, encodings, algorithms, fetch.txt, encrypted archives and special entries are unsupported. The Actor never downloads missing payload files. Percent/CR/LF filenames or target tokens, surrounding whitespace and leading-star target forms are unsupported because the pinned parser can alter their identity. Ordinary internal spaces and exact Unicode remain supported. Unsafe paths, duplicates, file/directory collisions and ambiguous Unicode names are refused. Exact nonambiguous Unicode names are retained. The scope does not include repair, profiles, certification or repository acceptance.

Results and charging

One dataset item contains schemaVersion, supportedScope, status, validationComplete, example, useful, reasonCode, findings, findingsOmitted, coverage, limits and provenance.

Status	Meaning	Useful paid event
valid	Complete supported validation found no integrity fault.	Productive runs only.
invalid	Diagnosed supported integrity faults, or malformed required structure.	Only complete productive validation.
unsupported	Scope, safety or resource limits prevent the promised evaluation.	No.
fetch_error	Secure acquisition failed or was refused.	No.

Missing/extra payload and checksum mismatches can coexist. Complete invalid reports retain all diagnosed faults; missing bytes are explicitly absent rather than reported as hashed. Findings sort by code and relative path. More than 100 findings are capped only after complete computation, with the exact omitted count.

For pay-per-event runs, the configured report-produced capacity is checked before productive acquisition. One event is charged only after a useful report is persisted. Persistence failure never charges. A rejected charge fails the run honestly after persistence; it is not retried. Separate invocations are separate requests. The initial experimental price is $0.05 per persisted complete productive valid or invalid report. Demo runs, refusals, fetch failures and incomplete validation do not charge a report event.

Private hosted QA produced the intended valid, checksum-mismatch and unsupported reports. Those owner tests do not establish customer demand or paid settlement. Full-limit latency/cost and commercial margin remain unmeasured. For large or private packages, use a local BagIt validator; this experimental service tests the convenience of a bounded hosted call.

Fixed limits

Compressed ZIP: 32 MiB (33,554,432 bytes). Expanded content: 64 MiB (67,108,864 bytes). ZIP entries including directories: 1,000. Paths: 1,024 characters. Parsed bagit.txt/bag-info/manifest/tagmanifest content: 8 MiB (8,388,608 bytes) cumulatively; ordinary unparsed tag files do not consume this additional ceiling. Distinct declared targets across algorithms: 1,000. Acquisition: 20 seconds. Aggregate acquisition/materialization/ validation: 120 seconds. At most five redirects and six requests; no retries. Limits are inclusive and cannot be raised through input.

Privacy and provenance

Process only packages you are authorized to inspect. Raw ZIP/payload bytes exist in temporary state and are removed after success, refusal, error or timeout. Reports contain source hostnames, ZIP SHA-256/byte count, request/redirect counts, validator version and authorized bag-relative filenames. They omit full URLs, source paths/queries, IP addresses, payload contents and raw exceptions. Apify retains the input URL and dataset according to its normal platform/caller retention settings; the Actor does not promise deletion of platform-managed input.

Development

Python 3.13, pinned Apify SDK, Pydantic v2 and bagit1.9.0 are locked with uv.

uv sync --locked --all-groups --python 3.13
uv run --locked ruff format --check .
uv run --locked ruff check .
uv run --locked pyright
uv run --locked pytest
uv run --locked pip-audit

Repository agents execute these workloads through rexec. Tests use synthetic fixtures and fake SDK/network boundaries, never public downloads or paid runs. See VALIDATION.md for observed evidence and remaining hosted unknowns. PRODUCT.md records the experimental product scope.