FAQ
Is the tool really free?
Yes, for the default path. Google Search Console, Yandex Webmaster, Yandex Metrika, Bing, CrUX and PageSpeed Insights all have free tiers, and no required dependency needs a card on file. Optional paid SERP and backlink adapters are documented but not shipped; if one is added later, it uses your own key and stays off by default. See Data sources and quotas.
Do I need a token for every provider?
Only for the providers you configure. GSC, Yandex Webmaster, Yandex Metrika, Bing and CrUX need a token or API key; PSI works without a key. lv24-seo auth status --json shows presence and source, never a value. See Configuration reference.
Does it scrape Google or Yandex?
No. Query, rank and indexing data comes from provider APIs. The project never scrapes SERPs, solves CAPTCHAs or automates Yandex Business. See Data sources and quotas.
Where is my data stored?
In the local SQLite database under storage.data_dir (data/seo.sqlite3 by default), plus report files you request with --out. A relative data_dir written in seo.toml is anchored to that file, so the same config uses the same database from any working directory; the built-in default and LV24_SEO_DATA_DIR remain process-scoped. There is no telemetry endpoint. Data leaves the machine only through provider API calls and the delivery channels you enable. See Security and privacy.
CLI or MCP — which should I use?
Both sit on the same engine and return the same data. The CLI is for terminals, cron and CI. The MCP server exposes 11 tools to coding agents over stdio or streamable HTTP. See MCP and agent skills.
Desktop app or CLI?
The desktop app wraps the same engine as a PyInstaller sidecar and adds a dashboard, settings forms and a recommendations page. The CLI remains the scriptable surface, and both read the same seo.toml and data directory. See Desktop app.
Does the desktop app update itself?
Yes, for Windows NSIS, macOS .app and Linux AppImage: the app checks latest.json, verifies a minisign signature and installs only after you confirm. deb, rpm and the portable ZIP are manual-only. Before the installer starts, the app stops only the engine sidecar PIDs it started; if one cannot be stopped, the update is blocked with UPDATE_ENGINE_BUSY and can be retried after that run exits. The 0.1.0 build predates the updater and needs one manual upgrade. See Releases and update.
Why are the builds unsigned?
OS code signing needs paid certificates and notarization, which this project does not buy. Windows and Linux packages are unsigned; macOS is ad-hoc signed. Updater payloads are separately minisign-signed, which protects updates but does not make the installer trusted by the OS. Verify SHA256SUMS.txt before running an installer. See Security and privacy.
Windows SmartScreen blocks the installer. What should I do?
SmartScreen, Smart App Control or antivirus may warn because the installer has no Authenticode signature. Download only from the release page, verify the SHA-256 against SHA256SUMS.txt, and let Windows run this specific app only if you accept the risk. Keep SmartScreen and antivirus enabled. See Install.
macOS refuses to open the app. What now?
The app is ad-hoc signed, so Gatekeeper warns on first launch. Open it once through System Settings → Privacy & Security → Open Anyway, or right-click the app and choose Open. Do not disable Gatekeeper. See Install.
How are tokens stored on Linux?
Through the OS Secret Service keychain when the engine's secrets extra is installed; the desktop passes the session-bus and XDG runtime variables the keychain needs. Without a keychain, set LV24_SEO_ALLOW_FILE_SECRETS=1 to use the 0600 file fallback, which prints a warning on every run. See Configuration reference.
Can I track several sites?
Not in one config: one seo.toml describes one project and one set of provider targets. Run one config per site and pass --config path/to/seo.toml; the desktop works with one selected config at a time. See Configuration reference.
Why do Yandex and GSC numbers differ?
They measure different things. GSC returns up to 50,000 rows per day with 2–3 days of latency. Yandex returns at most 3,000 queries per period, has no query-by-page join and lags about four days. The tool never sums across engines or fills missing rows with zero. See Data sources and quotas.
How do I change the report language?
Set report.locale to ru, en or zh (default ru). It controls email, Telegram, the CLI human summary and the HTML written by report weekly --out. JSON envelopes, snapshots, report.json and report diff stay canonical English. See Configuration reference.
How long is data kept?
Snapshots and crawl rows stay in SQLite until you delete the data directory; one snapshot is kept per kind, property and period. storage.raw_retention_days is reserved and currently a no-op because no raw archives are written. Back up the database before upgrading. See Security and privacy.
Why is my weekly report incomplete?
Providers lag and cap data. Check data_age_days, truncated, limitations[] and warnings[] in the envelope, and raise --lag-days if the newest days are still filling. A GSC final page that exactly fills the 50,000-row cap is reported as truncated, not complete. report diff also needs two stored weekly runs. See Troubleshooting.
Can I get logs?
Yes, but logging is opt-in. Set LV24_SEO_LOG_LEVEL=INFO (or DEBUG, WARNING, …) for the CLI or lv24-seo mcp serve; the redacting handler is installed only at those entry points. When stderr is not a terminal, each record is one JSON line. The --json envelope remains the stable diagnostic interface. See Configuration reference.
Can I run it in CI?
Yes. Use --json with exit codes, and report diff --markdown for a pull-request comment or job summary; warnings go to stderr so stdout stays valid Markdown. Keep tokens in CI secrets, never in seo.toml. See CLI reference.
How do I contribute?
Open an issue first, branch from main, keep one area per change, and run the full gate before a pull request:
uv sync --extra dev
uv run ruff check . && uv run ruff format --check . && uv run mypy src && uv run pytest -qSee the contributing guide for accepted changes and the commit format.
Where do I report bugs?
Engine, CLI and MCP issues: lv24-seo-optimizer issues. Desktop app issues: lv24-seo-optimizer-desktop issues. Include the lv24-seo version output, the exact command, the --json envelope and the exit code. Redact tokens and customer data.
Is any telemetry sent?
No. There is no telemetry endpoint and no [privacy] section to disable. The tool sends data only to the providers you query and the delivery channels you enable; the desktop contacts GitHub only for update metadata.
Source: _source/engine-cli.md, _source/engine-config.md, _source/engine-data-sources.md, _source/engine-security.md, _source/engine-releases-0.5.2.md, _source/desktop-README.md, _source/desktop-security.md, _source/desktop-releases-0.2.2.md.