One specialist agent per domain. Persistent memory, structured deliverables, sharper the more you use them.
An AI engineer who reads the code for you. Compare AI tools, decode hot projects, find alternatives, plan migrations — without reading source yourself.
Deep Research AI on demand. Industry reports, market analysis, competitive teardowns, and AI news — structured deliverables, not chat transcripts.
Generative Engine Optimization for the AI search era. Audit, optimize, and track what ChatGPT, Perplexity, and Google AI Overviews say about your brand.
Every specialist runs with persistent memory of your projects, your past asks, and the way you work. Each new run is sharper than the last.
Hand off a question and come back to a finished artifact. Each specialist scopes the task, gathers sources, reasons across them, and structures the result.
Repo wikis, research reports, GEO audits — every run produces a structured artifact you can share, edit, and re-run. Not a chat transcript.
A specialist runs end-to-end jobs in one domain — read a whole repo, run a multi-source research pass, audit a site for GEO. It scopes the task, fetches the data, reasons across sources, and returns a structured deliverable. Each specialist has persistent memory of your projects and gets sharper the more you use it.
Agents live in the same workspace, so context flows naturally — an AI Code Research investigation can feed into an AI Research market analysis, or an AI Research report can feed into a GEO content brief. You stay in one place; the agents coordinate behind the scenes.
Full agents. A HowWorks specialist runs autonomous jobs in the background — scoping the task, fetching data, reasoning across many sources, and returning a structured artifact. You can hand off a question and come back to a finished report.
Today HowWorks runs on a managed model selection optimized per agent — different specialists use different models depending on the task. Bring-your-own-key and bring-your-own-model are on the roadmap for teams with strict provider preferences.
Public-repo analyses are fair game for our public DeepDive index (you'll see your repo's analysis on /deepdive). Private content stays in your workspace and is not used to train any model. Full details in our privacy policy.
Join our early beta and get 300 free credits to explore products and run deep research. No credit card required.