Find Verified Jobs with Firecrawl

Use Claude and Firecrawl to find and verify live job openings.

Introduction

30 Second Summary

A job listing can look current in search results long after the original posting has changed or disappeared. Missing details make promising roles difficult to compare.

In this project, you will use Claude or ChatGPT with the Firecrawl integration to find live openings for a role and location you choose. You will verify the strongest matches against their original posting pages before ranking them against your criteria.

What You'll Build

Picture showing someone a ranked job shortlist where every opening links to an original posting that Firecrawl checked for current details.

By the end of this project, you'll have:

A live search table with up to eight openings for your chosen role and location. Every result includes a public source link.
A verified shortlist of up to three openings where page scraping confirms available details. Missing fields are clearly marked as Not stated.
A ranked top three based on two criteria you choose. Each role includes an evidence-backed reason plus a possible gap.
Secret Mission: Combine discovery plus verification into one reusable prompt for future job searches.

Are there any prerequisites?

You need a Claude or ChatGPT account plus a target role or job family. No coding or local installation is required because both supported paths run in your browser.

The Gemini chat website does not currently offer the same direct Firecrawl connection. Gemini developer integrations require code, API keys, or agent tooling, so they are outside this beginner workflow.

Before We Start

This is your moment to define what a useful job opening looks like for you. A clear goal gives you real criteria for judging every result later.

Connect Firecrawl to Claude

Connect Firecrawl to Your AI Assistant

Your job-search goal gives you a clear way to judge every opening you find.

Claude and ChatGPT can use Firecrawl only after you connect its browser integration. This step creates that connection without local software or API-key copying.

In this step, get ready to:

Choose Claude or ChatGPT for the project.
Connect the official Firecrawl integration through a selected Firecrawl team.
Enable Firecrawl in a new conversation.

Connect Firecrawl to your assistant

Claude provides Firecrawl as a connector, while ChatGPT provides it as an official plugin. Choose the assistant you want to use for the rest of the project.

Claude

Before you create a Claude account

Claude requires new account holders to be at least 18 and physically located in a supported location.

Phone verification can feel personal. Claude says it uses your number only for account verification.

Use an SMS-capable mobile number from a supported location. Claude does not accept VoIP numbers or landlines.

Choose a number you expect to keep. Claude does not let you change it after verification.

Choose the account path that matches your current access.

✔️ I have a Claude account

Open Claude in your browser.
Sign in with Google or your Claude email link.

ⓧ I need a Claude account

Open Claude in your browser.
Authenticate with Google or select Continue with email.
Open the email with the subject Secure link to log in to Claude.ai if you chose email.
Enter your eligible mobile number when Claude asks for verification.
Enter the six-digit code sent by text message.
Select Verify code.

Select Customize.
Choose Connectors.
Open Discover.
Select Firecrawl.
Select Connect.

You will see Firecrawl's authorization screen. This starts the shared account flow in the next substep.

ChatGPT

Open ChatGPT in your browser.
Sign in to your ChatGPT account.
Open Plugins.
Search for Firecrawl.
Open the official Firecrawl listing.
Select Install.

ChatGPT will open Firecrawl's authorization screen in your browser. This starts the shared account flow in the next substep.

Authorize the Firecrawl integration

Account authorization can feel sensitive even when your API key stays private. Firecrawl handles access through your account and the team you select.

Firecrawl's Free Plan requires no credit card. Integration requests use the plan of the Firecrawl team you select.

The authorization screen now follows one of two account paths.

✔️ I have a Firecrawl account

Sign in to Firecrawl.
Choose the Firecrawl team you want the integration to use.
Approve access for your selected AI assistant.

ⓧ I need a Firecrawl account

Create a free Firecrawl account from the authorization page.
Choose the Firecrawl team you want the integration to use.
Approve access for your selected AI assistant.

That sensitive setup is complete. Firecrawl is now authorized through your selected team.

Claude connection blocked by your organization?

This issue applies to Claude Team and Enterprise organizations. An Owner or Primary Owner may need to enable connectors before you can continue.

Ask an Owner or Primary Owner to enable connectors for your Claude organization.
Retry the Firecrawl connection after approval.

Still blocked? Help me troubleshoot Firecrawl connector approval in Claude.

Enable Firecrawl in a new conversation

Authorization makes Firecrawl available to your assistant. Start a new conversation so the integration loads with a clean chat.

Claude

Start a new Claude conversation.
Select +.
Open Connectors.
Toggle Firecrawl on for this conversation.

ChatGPT

Start a new ChatGPT conversation.
Open Plugins.
Confirm Firecrawl is installed and enabled.
Close the plugin menu.

Before you reopen the integration menu, do you think Firecrawl will still show as enabled in this conversation?

Claude

Select + again.
Open Connectors.
Confirm Firecrawl remains toggled on.

ChatGPT

Open Plugins again.
Confirm Firecrawl remains enabled.

You will see Firecrawl enabled for the new conversation. That confirms your assistant can use the integration.

Firecrawl not enabled?

Reconnect Firecrawl if Claude no longer shows it under Connectors.
Reopen Firecrawl under ChatGPT's Plugins if it is installed but disabled.
Start another new conversation after reconnecting or re-enabling Firecrawl.

Need help? Help me enable Firecrawl in my new conversation.

Your conversation is ready for live web research. Next, you'll use Firecrawl to discover current-looking openings without treating search snippets as verified records.

Discover Live Openings

Firecrawl integration is ready in your selected AI assistant. You can now use live web search to look beyond the assistant's existing knowledge.

Discovery gives you a fast set of job candidates. The important question is whether those first-pass results contain enough evidence to compare openings confidently.

In this step, get ready to:

Run a live search for openings that match your role and location.
Select three public employer or applicant-tracking-system sources.
Mark fields that need deeper verification.

Search for current openings

A focused search prompt gives Firecrawl clear boundaries for discovery. It also stops your AI assistant from filling gaps with assumptions.

Before you search, do you think every result will contain enough detail to compare the openings confidently?

Prepare the discovery request by copying the prompt below into the conversation where you enabled Firecrawl:

Use Firecrawl web search to find up to 8 currently advertised [[ROLE="your target role or job family"]] openings in [[LOCATION_OR_REMOTE="your target location or remote preference"]]. Prioritize public employer career pages and public applicant-tracking-system postings. Return title, company, location, source URL, and any posting date or status shown. Do not infer missing fields.

How does this prompt shape the search?

The first sentence tells your AI assistant to use Firecrawl for current job discovery.
The source rule focuses the search on public employer pages and public applicant-tracking-system postings.
The requested columns make the results easier to compare.
The final sentence protects the table from guessed information.

Replace the role placeholder with the target role or job family you chose earlier.
Replace the location placeholder with your target location or remote preference.
Send the completed prompt to start the Firecrawl search.

Your assistant should return a discovery table with up to eight rows. You should see a source URL on each usable result.

Why Is the Table Incomplete?

Firecrawl Search returns titles, descriptions, and URLs from its search results. Those results help you discover possible openings quickly.

The table may still contain blank or uncertain fields because a search result is only first-pass evidence. That visible shortfall is the intended result of this discovery step.

No table or source URLs?

Open Connectors in Claude or Plugins in ChatGPT.
Confirm Firecrawl is enabled in the current conversation.
Send the completed search prompt again.

Still missing live results? Help me troubleshoot the missing Firecrawl results.

Choose three public source pages

A trustworthy shortlist starts with sources you can inspect directly. Keep links from employer career pages or public applicant-tracking-system postings.

Read the source URL in each promising row.
Open the first promising source URL in a new browser tab.
Confirm the page shows the named opening.
Confirm the page does not require a login.
Confirm the page belongs to an employer or a public applicant-tracking system.
Repeat the source check until three URLs pass.
Exclude aggregator pages or generic pages that do not show the named opening.
Copy the three passing URLs into the checkpoint below.

Fewer than three public sources?

Return to the search prompt in your AI assistant conversation.
Remove one overly narrow modifier from your role phrase.
Expand a city location to its wider region when that still matches your goal.
Send the revised search prompt again.

Avoid using login-only pages to reach three results. Help me broaden my job-search phrase while keeping it aligned with my original goal.

Mark the discovery gaps

Discovery fields are clues that still need source-level proof. Labeling the gaps now prevents an incomplete snippet from becoming an unsupported claim.

Return to the three selected rows in your discovery table.
Record every empty field for each selected URL as blank in the checkpoint below.
Record every truncated or unsupported field as uncertain.
Record every field that conflicts within its row as inconsistent.

Before you check your work, do you think the discovery table now separates promising links from verified job records?

Count the public source URLs you selected.
Review the marked fields for each selected row.

You should have exactly three public URLs. Each URL should have its blank, uncertain, or inconsistent discovery fields marked for verification.

That first evidence pass is complete. Next, you will use the original pages to verify the details and rank the strongest matches.

Verify the Best Matches

Your live discovery table narrowed your search to three public source pages. Those links look promising. Their snippets still leave important details unresolved.

Search snippets provide a useful first view. The original posting pages hold the evidence needed for a complete job record.

The Firecrawl integration now uses page scraping to read those original pages. Your AI assistant turns the source details into a verified shortlist you can rank with confidence.

In this step, get ready to:

Scrape the three selected original posting pages.
Compare the source-backed details with the discovery table.
Rank the remaining openings against your two must-have criteria.

Scrape the original posting pages

A scrape reads the content on each selected page. This gives your assistant direct evidence for fields that the discovery snippets left blank or uncertain.

Return to the AI assistant conversation from the last step.
Confirm Firecrawl remains enabled for this conversation.
Paste this verification prompt into the chat box:

Use Firecrawl page scraping on the three URLs below. For each posting, return title, company, location, employment type, posting date, salary only when explicitly stated, three required skills, apply URL, and page status as active posting, unavailable, or unclear based only on the scraped page. Use Not stated for missing fields. Cite every source. Exclude any URL that redirects to a generic careers page or no longer contains the named role.

What does this prompt do?

Page scraping reads each original posting. This supplies source-page evidence for the requested fields.
The value Not stated preserves information that the source page does not publish.
The page-status labels separate confirmed openings from uncertain or unavailable pages.
The exclusion rule removes generic careers pages or pages that no longer contain the named role.

Paste the three selected source URLs on separate lines directly below the prompt.

Before you send the prompt, which marked fields do you expect the original pages to resolve?

Send the message to start the three page scrapes.

You'll see a cited table containing the source-backed details for each retained opening. Every missing source-page value should appear as Not stated.

You may see fewer than three retained rows if a selected URL redirects to a generic careers page. The same applies when a page no longer contains the named role.

That resolves the uncertainty in your discovery results. Your shortlist now comes from the original posting pages.

Scraped table missing or incomplete?

Confirm that Firecrawl is still enabled for this conversation.
Check that each pasted URL opens a public employer page or public applicant-tracking-system posting without a login.
Retry the scrape with one selected URL if your assistant does not process all three pages.

Still missing source details? Help me troubleshoot an incomplete Firecrawl job-page scrape.

Compare discovery with source evidence

Source verification compares the discovery snapshot with the original page. Each difference shows whether a field became confirmed, remained unpublished, or belonged to an invalid result.

Scroll back to the discovery table from the previous step.
Compare the title for each selected opening with its scraped row.
Compare the company for each selected opening with its scraped row.
Compare the location for each selected opening with its scraped row.
Identify every marked field that now has source-page evidence.
Identify every requested field that still shows Not stated.
Identify any selected opening that the scraped result excluded.

How should you read the differences?

A confirmed field now has direct support from the original posting. Keep the source citation attached so you can trace the detail later.

A Not stated value records the source page honestly. An excluded row prevents a generic or expired page from entering your final ranking.

Rank the verified openings

An evidence-based ranking uses only facts found on the scraped pages. Your two must-have criteria now decide the order without filling gaps with assumptions.

Paste this ranking prompt into the same AI assistant conversation:

Rank the remaining verified openings against my two must-have criteria from earlier in this conversation. Return the ranked table with title, company, location, employment type, posting date, salary only when explicitly stated, three required skills, apply URL, source URL, and page status. Keep page status as active posting, unavailable, or unclear. Keep Not stated for missing fields. For each role, add one evidence-backed fit reason and one possible gap. Use only the scraped source evidence. Do not invent qualifications or benefits.

How does this protect the ranking?

The ranking uses the two criteria you defined earlier in the conversation.
The field list preserves the details needed to inspect each opening.
The source URL keeps every ranked result traceable to its evidence.
The evidence rules prevent missing qualifications or benefits from becoming invented claims.

Before you send this prompt, which verified opening do you expect to rank first?

Send the message to generate the final ranking.

You'll see up to three openings in ranked order. Each row should preserve the verified job details with its source URL.

Each role should include one evidence-backed fit reason. It should also include one possible gap.

Confirm that every retained row includes an apply URL.
Confirm that every retained row includes a source URL.
Confirm that every retained row includes a page status.
Confirm that every absent source detail remains Not stated.
Confirm that each role has one evidence-backed fit reason.
Confirm that each role has one possible gap.
Confirm that excluded generic or outdated pages are absent from the ranking.

Ranking includes unsupported claims?

Ask your AI assistant to regenerate the ranking using only the scraped source evidence.
Check every fit reason against the cited posting details.
Replace any unsupported detail with Not stated.

Need help auditing the ranking? Help me find unsupported claims in my verified job shortlist.

You now have a shortlist grounded in the original job postings. Every ranking decision can be traced to visible source evidence.

Secret mission

Make the Workflow Reusable

Turn your verified job search into a reusable prompt. It discovers new openings before scraping the best three for ranking. You will test it with new criteria before auditing one result against its original posting.

Clean Up Your Resources

This browser-based workflow leaves no local processes running, so there are no ongoing local costs. Future requests through the Firecrawl integration consume credits from the selected Firecrawl team, so choose whether to keep the connection available, pause it in this conversation, or disconnect it completely.

Resources you used:

An authorized connection between your selected AI assistant and the Firecrawl team.
Firecrawl enabled in the conversation you used for the project.

Keep everything running

No action is needed. Choose this if you want Firecrawl ready for future job searches.

Leave the Firecrawl integration authorized in your selected AI assistant.
Enable Firecrawl only in conversations that need live web data.
Keep the verified shortlist in your current conversation.
Keep the reusable search-then-scrape prompt in your chosen note.

Pause - I'll come back to this later

Disable Firecrawl to prevent accidental requests. The authorized connection remains available when you return.

Claude

Return to the Claude conversation from earlier.
Select + beside the message field.
Open Connectors.
Toggle Firecrawl off for this conversation.
Confirm Firecrawl now shows as off.

ChatGPT

Return to ChatGPT.
Open Plugins.
Find Firecrawl.
Turn its enabled setting off.
Confirm Firecrawl now shows as disabled.

Delete - I don't want to use this again

Remove the Firecrawl integration if you no longer plan to use it. This blocks future requests through that connection.

Your results stay available

Disconnecting Firecrawl does not erase your verified shortlist. Your reusable prompt also stays in the note where you saved it.

Claude

Return to Claude in your browser.
Open Customize.
Choose Connectors.
Find Firecrawl in the connector list.
Disconnect Firecrawl from Claude.
Open Firecrawl's MCP settings.
Revoke the Claude connection.
Confirm Firecrawl no longer appears connected in Claude.

ChatGPT

Return to ChatGPT.
Open Plugins.
Find Firecrawl.
Remove the Firecrawl plugin.
Open Firecrawl's MCP settings.
Revoke the ChatGPT connection.
Confirm Firecrawl no longer appears in ChatGPT's installed plugins.

Nice Work!

You did it! You built a source-backed job-search workflow in Claude or ChatGPT with the Firecrawl integration.

You've learned how to:

Connected live web research to Claude or ChatGPT without writing code or handling an API key.
Used search-then-scrape verification to discover public job postings. Checked the original pages before treating their details as evidence.
Produced a cited and ranked shortlist based on your must-have criteria. Preserved missing details as Not stated instead of inventing information.
Secret Mission: Built a reusable prompt that searches before scraping selected source pages. Tested the workflow with a new role or location. Spot-checked one result against its visible source posting.

Ready to quiz yourself?