# Competitive research

Source: https://corespeed.io/use-cases/competitive-research

Competitive research with an agent is two steps: find sources, then read the few that matter. Your agent searches the web with web__search, reads pricing pages and changelogs with web__scrape, and writes a one-page brief to Notion with a source on every claim. No search provider account is needed; each call is priced in credits.

The job: Find competitor sources with web search, read only the pages that matter, and write a one-page brief to Notion.

## The prompt

Paste this line into Claude Code, Codex or Cursor:

```text
set up https://corespeed.io/SKILL.md (read it raw, e.g. with curl), then research our three closest competitors: find their pricing, recent launches and docs with web search, read only the pages that matter, and write a one-page brief to Notion with a source link on every claim.
```

It starts with the setup line, so it works on a machine without CoreSpeed: the agent fetches https://corespeed.io/SKILL.md, adds the MCP server and signs you in, then does the job.

## Set up first

Nothing for the research: no search provider account, and each call is priced in the catalog. When the brief is ready, your agent asks you to connect Notion if it is not connected yet.

## The tools it calls

- `web__search`: Find sources: pricing, launches, docs
- `web__scrape`: Read the pages that matter
- `web__google_search`: See how each one ranks on Google
- `notion__create_page`: Write the brief to Notion
- `memory__remember`: Save what changed, for next quarter

## Steps

1. **Name your competitors**: Paste the prompt into Claude Code, Codex or Cursor. Name the three, or let it find them.
2. **Let it search, then read**: Search returns sources, not answers. It reads only the few pages that matter.
3. **Review the brief**: Every claim links its source. Check the ones you will act on.

The full write-up: https://corespeed.io/blog/competitive-research-with-search-and-scrape

## Questions

### Does web__search write the answer?

No. It returns ranked sources with snippets. Your agent reads them, scrapes the ones that matter, and writes the brief itself.

### Can the agent crawl a competitor's whole site?

No. web__scrape reads one page per call, and there is no multi-page crawl. Search for the pages you want, then read each one.

### What does a run cost?

A search is 14 credits for up to 10 results, and a page read is 5 credits. A failed request is never charged.

### Do I need a search provider account?

No. Web search and reading are built into CoreSpeed. Each call comes out of your one credit balance.

Every tool and price: https://corespeed.io/catalog/web

Every tool and price: https://corespeed.io/catalog/notion

All use cases: https://corespeed.io/use-cases (Markdown: https://corespeed.io/use-cases.md)

## Related pages

- [What is an MCP gateway? One server for every tool](https://corespeed.io/glossary/mcp-gateway): An MCP gateway puts many tools and app accounts behind one remote MCP server and one sign-in. What it does, and what CoreSpeed's gateway holds for you.
- [Notion for Claude Code, Codex, and Cursor](https://corespeed.io/integrations/notion): Bring Notion project notes and working documents into the same conversation as the task at hand.

## Sources

- Web: https://corespeed.io/docs/web
- Connectors: https://corespeed.io/docs/connectors
- Competitive research: search for sources, then read what matters: https://corespeed.io/blog/competitive-research-with-search-and-scrape

Reviewed 2026-11-10.

## Links

- Product overview: https://corespeed.io/ (Markdown: https://corespeed.io/index.md)
- Documentation: https://corespeed.io/docs
- Pricing: https://corespeed.io/pricing (Markdown: https://corespeed.io/pricing.md)