AI prompt library for ChatGPT and Claude
501 prompts in 13 categories, from cold email to web scraping. Each one is written to run on live web data, so the answer comes back with sources instead of guesses.
Browse prompts by category
- GTM promptsICP research, positioning, pricing, launches, and market maps.65 prompts
- Coding promptsBuild scrapers, agent tools, and brand features into your app.53 prompts
- Lead Generation promptsProspect lists, decision makers, buying signals, and enrichment.48 prompts
- Cold Email promptsResearched first lines, sequences, follow-ups, and reply handling.36 prompts
- Web Scraping promptsExtract products, prices, tables, listings, and whole sites.46 prompts
- Web Search promptsDeep research, fact-checking, sourcing, and comparisons with citations.33 prompts
- File Parsing promptsPDFs, contracts, invoices, spreadsheets, slides, and scans.28 prompts
- News promptsCompany news, competitor alerts, briefings, and media monitoring.31 prompts
- Ad Copy promptsAd copy, hooks, audiences, landing pages, and competitor messaging.27 prompts
- Creative & Design promptsOn-brand image prompts, creative briefs, mockups, and brand kits.26 prompts
- Competitor Analysis promptsPricing, features, reviews, hiring, and change tracking.41 prompts
- SEO promptsSite audits, content briefs, search results research, and internal links.38 prompts
- Recruiting promptsCandidate sourcing, research, pay benchmarks, and outreach.29 prompts
How to use these prompts
Install the Context.dev MCP
Add the Context.dev MCP server to Claude, Cursor, or another agent, or install the ChatGPT app. This gives the assistant live web search, scraping, and data tools.
Pick a prompt
Copy it, or open it straight in Claude or ChatGPT. Every copied prompt tells the assistant to use Context.dev, links the MCP install page, and asks it to cite sources.
Fill in the brackets
Replace every [bracket] with your details, such as a company domain, a competitor, or a file, then run it.
Search every prompt
501 of 501 prompts
Build my ICP from closed-won deals
Here are my last [20] closed-won customers: [list of domains]. Read each website and find what they have in common: industry, size, business model, tech signals, and the trigger that likely made them buy. Turn it into an ICP table with firmographic criteria, buying triggers, and disqualifiers, naming the customers behind each pattern.
Uses: Scrape, Brand, Web Search. Install the Context.dev MCP if you haven’t already.
Add the SDK to my repo
Add Context.dev to [repo path] using the official SDK for [language]. Read https://docs.context.dev/sdks first, create one shared client that reads the key from the CONTEXT_DEV_API_KEY environment variable, and add a smoke test that scrapes [URL] to Markdown. Done means the test passes, the variable is in .env.example and the README, and no unrelated files changed.
Uses: Scrape. Install the Context.dev MCP if you haven’t already.
25 accounts that match my ICP
Find 25 companies that match: [industry], [employee range], [country], and [signal, such as “uses Shopify” or “hiring SDRs”]. Return a table with name, website, size, location, and the evidence for each criterion.
Uses: Company Search, Web Search, Scrape. Install the Context.dev MCP if you haven’t already.
Research first, then write
Research [prospect name] at [company domain]: their role, recent public posts, company news, and hiring. Then write a cold email under 90 words offering [our offer]. The first line must reference something specific you found, with the source link below the email.
Uses: People Enrich, Scrape, News Search. Install the Context.dev MCP if you haven’t already.
Product page to JSON
Scrape [product URL] and return JSON with name, brand, price, currency, availability, variants, rating, review count, and image URLs. Use null for any field that isn’t on the page.
Uses: Scrape. Install the Context.dev MCP if you haven’t already.
Deep research report
Research [topic] in depth. Break it into the five questions an expert would ask, search at least 15 sources, read the important ones in full, and write a report with an executive summary, findings per question with a confidence level, disagreements between sources, and a dated source list.
Uses: Web Search, Answers, Scrape. Install the Context.dev MCP if you haven’t already.
Contract key terms
Parse this contract: [file]. Extract parties, term, renewal, termination rights, payment terms, liability caps, indemnities, and governing law into a table with page references. Flag anything unusual.
Uses: Parse Files. Install the Context.dev MCP if you haven’t already.
Morning briefing
Give me a morning briefing on [industry or topics]: the five most important stories from the last 24 hours, why each matters to [my role], and links with publish times.
Uses: News Search, Web Search. Install the Context.dev MCP if you haven’t already.
Google search ads from my landing page
Read [landing page URL]. Write 15 Google responsive search ad headlines (30 characters max) and 4 descriptions (90 characters max) that match the page’s offer. Show the character count for each.
Uses: Scrape. Install the Context.dev MCP if you haven’t already.
Brand kit from a URL
Pull the logos, color palette, fonts, and visual style from [website]. Build a one-page brand kit with hex codes, type pairings, logo usage notes, and three example layouts.
Uses: Brand, Styleguide. Install the Context.dev MCP if you haven’t already.
Full competitor profile
Build a profile of [competitor domain]: product, target customer, pricing, positioning, team size, recent news, hiring, reviews, and strengths and weaknesses, with a source for each.
Uses: Brand, Scrape, News Search, Web Search. Install the Context.dev MCP if you haven’t already.
Content brief from the top results
Search [keyword] and read the top ten results in full. Write a content brief with search intent, the subtopics every top page covers, gaps none of them cover, a suggested outline, and a target length.
Uses: Web Search, Scrape. Install the Context.dev MCP if you haven’t already.
Source a shortlist
Find 20 people who could fill this role: [job description]. Match on [must-have skills], [years of experience], and [location or remote]. Give name, current title and company, profile link, and why each one fits.
Uses: People Search. Install the Context.dev MCP if you haven’t already.
Positioning canvas with evidence
Work through April Dunford’s positioning steps for [product]: competitive alternatives (what buyers actually use today, from reviews and forums), our unique attributes (checked against [competitors]’ sites), the value they enable, best-fit customers, and market category. Link the evidence for each step, then write three positioning statements to test.
Uses: Web Search, Scrape. Install the Context.dev MCP if you haven’t already.
Interview me before building
I want to add [feature] to [app] with Context.dev. Read https://docs.context.dev, then interview me with short multiple-choice questions until you could write a spec: which endpoints, inputs and outputs, failure handling, credit budget, and what’s out of scope. Write the answers to SPEC.md and stop without writing code.
Uses: Web Search, Scrape, Answers. Install the Context.dev MCP if you haven’t already.
Decision makers at target accounts
For each of these companies [domains], find the person most likely to own [problem], plus one backup. Give name, title, profile link, and why they’re the right person.
Uses: People Search, Web Search. Install the Context.dev MCP if you haven’t already.
First lines for a whole list
For each company in this list [domains], find one recent, specific fact from their website, blog, or news, and write a first line that connects it to [our value prop]. Return a table with company, fact, source, and first line, and write “no hook found” rather than forcing a weak one.
Uses: Scrape, News Search, Web Search. Install the Context.dev MCP if you haven’t already.
Whole category to CSV
Scrape every product in the category at [category URL], following pagination. Output a CSV with name, URL, price, and stock status, and tell me how many pages you visited.
Uses: Scrape, Map URLs, Crawl & Batches. Install the Context.dev MCP if you haven’t already.
Primary sources only
Answer [question] using only primary sources: official sites, filings, papers, government data, or original announcements. Exclude aggregators and SEO blogs, and quote the relevant passage for each claim.
Uses: Web Search, Scrape. Install the Context.dev MCP if you haven’t already.
Invoices to JSON
Parse these invoices: [files]. Return JSON per invoice with vendor, invoice number, date, due date, line items with quantity and unit price, tax, and total. Check that the line items add up to the total.
Uses: Parse Files. Install the Context.dev MCP if you haven’t already.
Company news digest
Find all news about [company domain or ticker] from the last 30 days. Group it into product, people, financial, legal, and partnerships, with a one-line summary and link for each story.
Uses: News Search. Install the Context.dev MCP if you haven’t already.
Meta ads in customers’ words
Find reviews of [product] and similar products. Pull the exact phrases customers use about [pain or outcome] and write five Meta ad variations (primary text, headline, description) built on that language.
Uses: Web Search, Scrape. Install the Context.dev MCP if you haven’t already.
On-brand image prompts
Read [brand domain]’s colors, imagery style, and tone. Write five detailed image-generation prompts for [campaign] that use the exact palette and match the existing photography style.
Uses: Brand, Styleguide, Scrape. Install the Context.dev MCP if you haven’t already.
Feature comparison matrix
Read the product pages and docs of [competitors] and [our site]. Build a feature matrix for [feature list] marking each as supported, partial, or missing, with a link to the evidence for each cell.
Uses: Scrape, Crawl & Batches. Install the Context.dev MCP if you haven’t already.
Showing 24 of 501
FAQ
What is an AI prompt library?
A collection of ready-to-use prompts you can paste into ChatGPT, Claude, or another assistant instead of writing one from scratch. This library has 501 prompts in 13 categories, each written for work that needs current information from the web.
Do these prompts work in ChatGPT and Claude?
Yes. They work in ChatGPT, Claude, Cursor, and any assistant that accepts a text prompt. Every card can copy the prompt or open it directly in Claude or ChatGPT.
Why install the Context.dev MCP before running a prompt?
Most of these prompts ask the assistant to search the web, open pages, read files, or find people and companies. The Context.dev MCP gives the assistant those tools, so it works from current sources instead of its training data. Every copied prompt links the install page in case it isn’t set up yet.
Are the prompts free to use?
Yes. Copy, edit, and reuse them however you like. Running them with Context uses credits from your Context plan, as listed on the pricing page.
Start building with live web context.
Scrape, search, and research the web with one API. Bring the results straight into your product.