Crawllr
Crawling is a solved problem. Crawling in a way an LLM can actually use is not.
- Type
- Tooling
- Timeline
- 2025 — present
- Role
- Solo — architecture and build
The problem
Fetching pages is trivial. Turning a site into something a language model can reason over is not. Boilerplate has to go, structure has to survive, and the result has to be chunked in a way that keeps meaning intact across boundaries.
Most crawlers optimise for coverage. This one optimises for what comes after.
What it does
Large-scale collection with knowledge extraction built into the pipeline rather than bolted on afterwards. Output is indexed for retrieval, not just archived.
It is exposed to agents over MCP, so an agent can crawl and reason in one loop instead of waiting on a separate ingestion step.
Where it is
Running in production, feeding the other projects on this list.
Skills applied
Engineering
- Crawl architecture
- MCP integration
- Vector storage
AI
- Knowledge extraction
- Content chunking
- Retrieval indexing
Built with
- PHP
- MCP
- LLM
- Vector storage
More work
Agentic Marketing Platform In development An autonomous digital marketing platform where collaborative agents run technical SEO, competitor research, content planning, local SEO, analytics, and reporting — the full process, run by agents. Agentic SaaS Tools Medellin.co 2020 — present An AI-powered city platform using RAG, intelligent search, autonomous content generation, and AI assistants for travelers, expats, and local businesses. Websites SaaS Agentic Verificaa.com 2021 — present ID verification and fraud prevention for Latin America — letting people verify businesses and individuals before doing business with them. SaaS Tools Agentic