Genspark
Consumer general-purpose 'super agent' / AI workspace blending multi-LLM routing, browser actions, code execution, generated 'Sparkpages' and downstream apps; built and operated by MainFunc.
Description
Genspark markets itself as an AI-powered workspace whose Super Agent autonomously completes user tasks — from research to running multi-step workflows. The company's GitHub publicly showcases Clawverse, claimed to be built by 100+ autonomous coding sprints of 17,500 lines with zero human-written code. Public, citable technical documentation for Genspark's internal agent loop is scarce: most claims about a mixture-of-agents (MoA) or specific tool counts circulate through press/marketing rather than verifiable docs. The product page and several deep technical pages return 403 to unauthenticated scrapers, so several claims are recorded here at evidence_status='none'.
Solution
Hosted consumer agent loop; not documented in citable upstream form. From public surfaces and the GitHub showcase: a user task is decomposed into sub-actions executed against multiple LLMs, a browser surface and code-execution capability; results are aggregated into long-form 'Sparkpages' or app artifacts. Internal details (which models, what routing, what step budget) are not published.
Primary use cases
- consumer research and task agent
- generated 'Sparkpages' answer pages
- AI-built apps / artifacts (e.g. Clawverse)
- broad task automation across browser, search, code
Open the full interactive page →
Diagram, neighbourhood map, code examples, related patterns and full provenance.