cognium.dev is built with GSAP, hosted on Cloudflare. It measures traffic with Cloudflare Web Analytics. It sends email through Amazon SES.
| Technology | Version | First seen |
|---|---|---|
| JavaScript Frameworks & Libraries3 | ||
| – | 7 Nov 2026 | |
| – | 7 Nov 2026 | |
| – | 7 Nov 2026 | |
| Frontend Build Tools & CSS1 | ||
| – | 7 Nov 2026 | |
| CDN & Hosting1 | ||
| – | 7 Nov 2026 | |
| Analytics & Tag Management1 | ||
| – | 7 Nov 2026 | |
| Transactional Email2 | ||
| – | 7 Nov 2026 | |
| Fonts & Icon Fonts1 | ||
| – | 7 Nov 2026 | |
| Maps, Geocoding & Geolocation1 | ||
| – | 7 Nov 2026 | |
Registration from the registry's RDAP record, the certificate the site serves, and its rank in public traffic lists.
The crew hasn’t read this site’s colors yet. We read a palette when we capture a site’s screenshot, busiest sites first. Meanwhile, browse palettes from other sites.
mx.zoho.commx2.zoho.commx3.zoho.comnucum.ns.cloudflare.comwalt.ns.cloudflare.comv=spf1 include:zohomail.com ~all# cognium.dev robots.txt — same policy as cognium.net. # Indexable and quotable by search and answer engines; not licensed for model training. # If Cloudflare's managed robots.txt block is enabled for this zone, keep its signals in line with this file. User-agent: * Allow: / Content-Signal: search=yes, ai-input=yes, ai-train=no, use=reference # Answer engines that cite their sources are welcome. User-agent: OAI-SearchBot Allow: / User-agent: ChatGPT-User Allow: / User-agent: PerplexityBot Allow: / User-agent: Perplexity-User Allow: / User-agent: Claude-User Allow: / User-agent: Claude-SearchBot Allow: / User-agent: Applebot Allow: / User-agent: Bingbot Allow: / User-agent: Googlebot Allow: / # Training crawlers: declined, consistent with the signal above. User-agent: GPTBot Disallow: / User-agent: ClaudeBot Disallow: / User-agent: Google-Extended Disallow: / User-agent: Applebot-Extended Disallow: / User-agent: CCBot Disallow: / User-agent: meta-externalagent Disallow: / # LLM guidance: https://cognium.dev/llms.txt (index) and /llms-full.txt (full text) Sitemap: https://cognium.dev/sitemap.xml
# cognium.dev > Full page text: https://cognium.dev/llms-full.txt > Autonomous software delivery. Start with your existing repo: give Cognium a concrete task, its agent makes the change, Cognium evaluates the result, and you control what gets accepted. The open-source engine is available now; the Cognium Autonomous Loop is in early access for developers in November 2026, free for open source and individuals. cognium.dev is the developer home of the Cognium Autonomous Loop and of its open engine, cognium-dev (MIT licensed). ## What you can give it - Fix security findings: Cognium reports a finding, the agent proposes a fix, Cognium scans the changed code again. Endpoint: the finding is no longer reported on re-scan. Findings that appear for the first time on re-scan are reported as newly reported and routed to you. - Find unused code: Cognium finds a candidate through call-graph reachability; you inspect the references and decide whether to remove it. Endpoint: a candidate ready for review. - Fix accessibility issues: a supported accessibility check finds an issue, the agent changes the code, Cognium re-checks it. Endpoint: the check passes. ## How it works 1. Cognium supplies the finding and the requirement it relates to. 2. It coordinates your coding agent, or its own, on the task. 3. It evaluates the change deterministically against the requirement: same inputs, same result. 4. Unresolved findings go back to the agent with the reason, until the change passes or needs you. ## What you control You connect a repo and choose its workflows. For each workflow you set when it runs, how far the agent may go (propose only, open a pull request, or merge when the evaluation passes) and who approves. Rules are set in Cognium, once per workflow, and apply to every change it produces. The evaluation result and the approval are reported separately; a workflow merges on its own only if you set it to. The requirement, the check and the evaluation record stay with the change for the next one. Agents connect over MCP; Cognium hands out tasks over A2A. ## The open engine (available now) - Install: `npm install -g cognium-dev`, then `cognium-dev scan ./src`. Output includes SARIF 2.1.0. - MCP server: `npx -y @cognium/mcp-server` (13 tools, including scan, taint_paths, explain_finding, check_sanitizer, find_callers). Source: https://github.com/cogniumhq/cognium-mcp - Source and license: https://github.com/cogniumhq/cognium-dev (MIT) ## Evidence - CWE-Bench-Java, 120 CVEs, CWE class given to every tool: Cognium 107 (research mode), IRIS + GPT-4 55, CodeQL 27 (baselines as published by the benchmark's authors, Li et al., ICLR 2025, arXiv 2405.17238). - A scan of 300 repositories in 6 languages (Java, JavaScript and TypeScript, Python, C#, Rust, Go): 13 issues accepted upstream. - OWASP Benchmark Java v1.2, all 2,740 cases, one open scorer: cognium-dev 4.9.13 Youden 73.5% (TPR 91.0%, FPR 17.4%); CodeQL 2.27.0 59.9%; Semgrep OSS 1.177.0 47.9% (September 11, 2026). - Method and results: https://cognium.dev/benchmark · published artifacts: https://github.com/cogniumhq/sast-benchmarks ## Pages - [Home](https://cognium.dev/): what you can give it, a worked example, how it works, what you control, early access - [Benchmarks](https://cognium.dev/benchmark): method and results - [Documentation](https://cognium.dev/docs/): first scan, GitHub Actions, MCP setup - [Research](https://cognium.dev/articles/): OWASP Benchmark Java results; SAST false-positive rates explained - [Community](https://cognium.dev/community): discussions and contributing ## Related - Specifica, version-controlled specifications: https://specifica.org - Skills Registry, agent skills reachable over MCP: https://skillsregistry.net - Cognium: https://cognium.net
| – |
| 7 Nov 2026 |