Deepcrawl is called Lumar now. The rebrand is announced on lumar.io itself, so a comparison against Deepcrawl is a comparison against Lumar, and search results for the old name keep resolving to the new site. Everything below uses the current name.
The two tools sit at opposite ends of the same category. Screaming Frog SEO Spider is a desktop application installed and licensed per person. Lumar is a cloud platform sold to organizations. Both crawl websites and report technical problems. Almost everything else about them differs, including who signs off on the purchase.
There is a third shape the question usually misses, so the table below carries it as a column: a command-line binary that runs the crawl where your build runs. That is ours, squirrelscan, and the section near the end says plainly what it does not replace.
The short version
| Screaming Frog SEO Spider | Lumar | squirrelscan | |
|---|---|---|---|
| Delivery | Desktop app for Windows, macOS and Linux | Cloud platform, four apps | CLI binary for macOS, Linux and Windows, plus a hosted cloud |
| Products | One crawler, plus a separate Log File Analyser | Analyze, Monitor, Protect, Impact | One binary, squirrel, with an MCP server and agent skills |
| Free tier | Free version capped at 500 URLs per crawl | Not published | Local audits free signed out, to 10,000 pages; free cloud tier up to 500 pages per audit |
| Price | £199 per licence per year, per user | Not published, contact form | Free locally while signed out, and a free cloud tier; Pro $19 a month or $190 a year |
| Checks | Over 300 SEO issues | More than 250 reports in Analyze, more than 350 automated QA tests in Protect | 300+ audit rules across 21 categories |
| Accessibility | AXE engine, WCAG 2.0 A to AAA then 2.1 AA and 2.2 AA, enabled in config | WCAG 2.2, levels A to AAA | 61 accessibility rules, no conformance level claimed |
| CI/CD | Command line and scheduled crawls | Protect integrates with the pipeline and can block a build | Runs as itself in CI, --fail-on exits 2 on a threshold |
| MCP server | Since v24.0, paid licence, runs crawls on your machine | Not mentioned on lumar.io | Hosted and local (squirrel mcp) |
| Buying it | Card, today | Quote, procurement | Install and run, card for Pro or Team |
The Screaming Frog and Lumar rows were read from each vendor's own site, first on 10 September 2026 and re-checked on 23 October 2026. The squirrelscan rows come from this repository rather than a marketing page: the rule counts from the rule catalogue, the prices from the plan definitions, the flags from squirrel audit --help.
A note on the accessibility row, because it is the one most likely to mislead. Screaming Frog and Lumar both state a WCAG conformance level. squirrelscan does not, and the row says so: its 61 accessibility rules are hand-written checks, not a WCAG conformance engine, and reading that row as a smaller version of the other two would be reading it wrong.
Screaming Frog
The SEO Spider ships as a desktop application, with builds for Windows, macOS on both Apple Silicon and Intel, and Linux. You install it, point it at a domain, and a crawl fills a table you can filter, sort and export. The free version caps a crawl at 500 URLs. A licence costs £199 per user per year, dropping to £189 at five licences, £179 at ten and £169 at twenty or more. Licences are per user, so two people need two of them.
Coverage is broad: over 300 SEO issues, broken links, metadata, duplicate content, XML sitemap generation and site architecture visualisation. JavaScript sites crawl through an integrated Chromium rendering service, so Angular, React and Vue apps render rather than returning empty shells. Custom extraction pulls any field out of the HTML with XPath, CSS selectors or regex. Crawl comparison diffs one crawl against another. Integrations cover the Google Analytics, Search Console and PageSpeed Insights APIs, plus link metrics from Majestic, Ahrefs and Moz. Recent versions run custom AI prompts against pages mid-crawl, with OpenAI, Gemini, Ollama and Anthropic as the model options. Version 24.0, released 19 May 2026, added an MCP server, so Claude and other assistants can run crawls and work with the data on your machine. Like scheduling, it needs a paid licence.
Log file analysis is a separate purchase. Screaming Frog ships it as its own application, the Log File Analyser, downloaded and licensed apart from the SEO Spider. Lumar keeps its equivalent inside the platform.
The interaction model is what people pay for. Someone who knows the filters can answer a question about a crawl faster here than anywhere else in the category. The constraint is where the work runs. Scheduling and command-line launch parameters do automate a crawl, but on a machine you license, patch and keep available, and the artefact at the end is an export rather than a verdict a build step can read.
Lumar
Lumar splits the same job across four apps. Analyze is the crawler, with more than 250 built-in reports and custom extraction, built for enterprise-scale sites. Monitor tracks metrics across multiple domains with dashboards and alerts. Protect handles the pipeline case, with more than 350 built-in automated QA tests covering SEO, site speed and accessibility, integration with a CI/CD pipeline, the option to stop a build that misses a health threshold you set, and template testing on a schedule as often as hourly. Impact reports the value of optimization work back to stakeholders.
Accessibility sits closer than the positioning suggests. Both tools cover it. Screaming Frog runs the open-source AXE engine, grouping violations from WCAG 2.0 A up through 2.0 AAA and then 2.1 AA and 2.2 AA, switched on under Config, Spider, Extraction. Lumar's coverage reaches WCAG 2.2 across levels A to AAA and sits in the product rather than behind a setting. Read those two as the same ballpark rather than a gap. The wider difference is elsewhere: Lumar publishes GEO and AEO metrics for AI search visibility, a category that postdates both tools' original design.
Lumar also exposes an API for building on its data, which matters for the case the platform is sold into: reporting that has to land in a warehouse or a BI tool rather than stay in a vendor dashboard.
Pricing is a contact form with no tiers and no published numbers, alongside a note about flexible pricing and packaging. That is standard for the segment, and it identifies the buyer. This is a purchase with a budget line and a procurement process behind it.
The published check counts do not compare. Screaming Frog counts SEO issues it can flag on a page. Lumar counts reports in Analyze and automated tests in Protect, which are different units in two different products. Reading 300 against 250 as a coverage gap is reading a number that was never the same measurement.
Which one fits
Screaming Frog suits investigative work owned by one person: auditing a client site, working out why a section fell out of the index, extracting a field from ten thousand product pages. The price is known, the licence is yours today, and the filtering depth rewards familiarity with the tool.
Lumar suits crawls that have to run without anyone starting them, across more sites than a laptop holds, for a team that shares the results. Build-blocking in Protect and multi-domain monitoring are capabilities Screaming Frog does not offer, and past a certain scale the crawl itself outgrows a desktop machine.
squirrelscan suits the case where the crawl belongs in the repository: one site, checked on every deploy, by whoever is already editing the templates. It is a binary, so CI runs it without a licence file or a desktop, and a coding agent can run it without a person. Signed out it is free locally, which makes it the cheap way to answer "did that regress" between the deeper audits you do elsewhere. There is a hosted side with a dashboard and published reports, and a free tier of it, but the command line is the product and the rest is where a run goes afterwards.
Teams between the first two positions usually buy Screaming Frog and feel the gap on automation, or price Lumar and decide the scale does not justify it. Closing that gap with a CLI in CI is the cheaper move than going up a tier, and it leaves the desktop licence for the investigation work it is good at.
What happens after the report
Both tools assume a person reads the output. Screaming Frog produces a table. Lumar produces a dashboard, and in Protect's case a build gate someone configured. In both cases a person then reads the findings, decides what matters, opens the codebase and edits it. That translation step is the slow one, and it is the step a coding agent can take over, given output written for one.
squirrelscan covers that narrower job. It is a command-line binary, squirrel, that crawls a site and scores it against 300+ audit rules across SEO, performance, security, accessibility, content and agent experience. Signed out, local audits are free.
curl -fsSL https://install.squirrelscan.com | bash
squirrel audit https://example.com --format llmBeing a binary rather than an application has two consequences. It runs in CI as itself, and --fail-on 'score<90' exits non-zero when a threshold trips, so a build step reads the result directly. It also runs inside a coding agent. Point Claude Code, Cursor or Codex at the repo with the squirrelscan skill or MCP server installed, and the agent audits the site, reads the findings, edits the templates and re-runs the audit to confirm the fix landed.
It does not replace Lumar for an enterprise. A CLI has no multi-domain dashboard, no stakeholder reporting app, no procurement-friendly contract and no account team, and an organization that needs those needs a platform. It does not replace Screaming Frog for open-ended manual investigation either, because there is no table to filter. It covers the step where a person would otherwise retype findings into a code editor.
The Screaming Frog alternatives roundup covers the rest of the desktop crawler field including the free ones, and Sitebulb alternatives covers the closest competitor to Screaming Frog on the desktop. For the agent workflow, fixing audit issues with an AI agent walks through it on a real site.