find-all

# Find All **Claude can now query the public web like a database.** Type `/find-all`, then one sentence. It comes back as a table, with the link each cell came from. ``` /find-all every startup that shut down in 2026, how much it raised, and why it says it failed /find-all every open job at OpenAI and Anthropic, with the salary /find-all every SaaS that killed its free plan this year /find-all every public company that said "AI agents" on its last earnings call ``` ## Why you can trust the table - **One source per cell, not per row.** A cell carries the page it came from, or it stays empty. - **Blanks are listed, with the reason.** No "N/A", no guessed range. - **Weak claims get rejected, and you see why.** Asked for companies that replaced sales or support staff with AI in 2026, the AI layoff lists named dozens. Five held up at the source. The rest went to a REJECTED section: other years, other functions, anonymous sources, or companies that said outright AI was not the reason. - **When sources disagree, both ship.** Two headcounts, two links. It never picks one for you. - **Every table ends with a coverage line.** What was searched, what survived, what failed to load. "Find all" means all it found, never all that exist. ## You need Bright Data (free to start) Find All runs entirely on the Bright Data MCP. It is the skill's only dependency. **Bright Data gives every account 5,000 free requests a month, no credit card required, renewed on the 1st.** 👉 **[Create your free Bright Data account](https://brdta.com/toolmonsters)** Then copy your API token from your account settings (Settings → Users → API token). ## Setup ### 1. Connect Bright Data **Claude (web or desktop):** Settings → Connectors → Add custom connector, and paste: ``` https://mcp.brightdata.com/mcp?token=YOUR_API_TOKEN&groups=business&tools=search_dataset,list_dataset_fields ``` **Claude Code:** ```bash claude mcp add --transport http brightdata "https://mcp.brightdata.com/mcp?token=YOUR_API_TOKEN&groups=business&tools=search_dataset,list_dataset_fields" ``` The base tools (search and page reading) are always on. `search_dataset` and `list_dataset_fields` add the indexed company and people datasets. Without them the skill still works, in live search mode, and says so. ### 2. Install the skill **Claude (web or desktop):** download [`find-all.zip`](find-all.zip), then Settings → Capabilities → Skills → upload it. **Claude Code:** ```bash git clone https://github.com/ToolMonsters/find-all cp -R find-all/find-all ~/.claude/skills/ ``` ### 3. Ask for rows Type `/find-all` and describe the rows. Name the columns if you know them. Claude echoes back the mode, the filters and the columns before spending anything, so you fix the query before the run, not after. ## What you get back ``` Find All: "Series B fintechs in Germany hiring a Head of Compliance" Indexed mode · 3 dataset pages · checked 2026-08-18 1,294 matched the filter · 40 returned · 5 columns | Company | HQ | Headcount | Last round | Compliance role open | | Northwind | Berlin | 340 ¹ ² | Series B, Mar 2026 ³ | Yes ⁴ | | Kalder | Munich | 610 ¹ | | Yes ⁴ | | Vasari | Hamburg | 210 ¹ | Series B, Nov 2025 ³ | | ¹ company dataset record ² company_size reads "201-500", conflicts with 340 ³ crunchbase.com/organization/northwind ⁴ job listing URL BLANKS Kalder last round: no public funding record found. Vasari compliance role: no matching listing found. COVERAGE Filtered on industry, country and headcount. This is what the dataset holds, not the whole German fintech market. ``` ## How it works | The rows are... | Mode | How | |---|---|---| | Companies or people | **Indexed** | Bright Data's datasets with a real filter tree. Thousands of matches, paginated, instead of ten search results. | | Anything else | **Live** | 5 to 10 search angles at once, then the pages are read in full, never the snippets. | | A row points at a supported platform | **Enrich** | The matching Bright Data tool fills the missing fields for that row. | ## Good to know - **"Find all X that recently did Y" is the strongest shape.** The recent event narrows the set and is usually why you're asking. - **Ask for fewer columns than you want.** Five sourced columns beat twelve half empty. - **Re-run it next month and diff.** Who entered, who left, what changed. - **Watch the request count on broad questions.** A narrow indexed query takes a few dozen requests. A wide live question can read 100+ pages. The free tier covers plenty of runs, and Claude confirms the query before it spends. - **The datasets are a sample, not a census.** A result count is never a market size. - **Export it.** CSV or JSON, with a source column next to every field. ## What it does NOT do - Fill a cell without a source. - Claim the result is exhaustive. - Return email addresses. Bright Data does not have them. - Pick a side when sources disagree. - Use any tool other than Bright Data. --- Built by Yonathan Cohen · [Tool Monsters](https://toolmonsters.com) · hello@toolmonsters.com