How to Use the LinkedIn API in Your Agent

Get company data into your agent without copying LinkedIn pages. Paste this into your agent:

Set up https://monid.ai/SKILL.md, and then use Monid to inspect the LinkedIn company-profile lookup from TikHub. Get the company profiles for https://www.linkedin.com/company/microsoft/ and https://www.linkedin.com/company/google/. Return company name, company ID, followers, employees_in_linkedin, company_size, industry and headquarters. Keep source URLs and collection dates. Omit all personal data. Ask before extending the list.

Planning estimate: 1 min setup · 1-2 hrs saved per 50-company refresh.

If you are a developer building a research agent, a LinkedIn API search starts with a practical problem: 50 company URLs still mean 50 pages to open when your script has no data connection. Copying names and follower counts into a spreadsheet leaves the next refresh just as manual.

A browser bookmark does not give your program a structured record. A copied description also mixes company claims, workforce labels and numbers that your agent must keep separate before it summarizes anything.

This guide uses Monid, the OpenRouter for agent tools, to retrieve company profiles through TikHub. We'll inspect the input, execute a small lookup, and turn the response into a repeatable company research table.

Here is the finished sample from two calls on November 11, 2026. These are returned company-page values, with personal fields omitted.

How to use the LinkedIn API with Monid

Use Monid to discover a company lookup, inspect its required URL, and execute one request before connecting a list. An endpoint is a callable operation with its own input schema and billing rule.

This is third-party LinkedIn data retrieval through TikHub, not LinkedIn's official API. The demonstrated operation returns a company profile; it does not publish posts, send messages or manage your account.

Step 1. Install and add your key

For agents, get an API key at app.monid.ai, then paste the setup instruction and supply the key through your agent's credential configuration:

set up https://monid.ai/SKILL.md

For humans, install the CLI and register your key locally:

npm install -g @monid-ai/cli
monid keys add -k <your-api-key> -l main

Keep the credential out of prompts that you publish, source control and exported research files. The saved key authorizes paid calls, so your workflow needs a spending boundary before it loops over company URLs.

Step 2. Choose company data

Search the catalog, then inspect the TikHub company-profile lookup so the agent reads the actual input and price rather than guessing a route from the search phrase.

monid discover -q "linkedin api" -l 8 -j --include-unavailable
monid inspect -p tikhub -e /api/v1/linkedin/web_v2/get_company_profile

Discovery and inspection are free. Choose the company operation because the input list contains organizations; company and personal profile URLs belong to different jobs.

Step 3. Request one company

Use the full company URL in --query because inspection declares url as a query parameter, rather than a JSON body field.

monid run -p tikhub -e /api/v1/linkedin/web_v2/get_company_profile \
  --query '{"url":"https://www.linkedin.com/company/microsoft/"}' -w -j

The Microsoft request completed with one result and one billed unit. Its execution interval, from the recorded start to completion, was 5.359 seconds; the Google request took 23.975 seconds.

Step 4. Keep the response and its receipt

Retain the run ID alongside the company row so you can retrieve the completed result without issuing another lookup.

monid runs get -r 01M75P5780G9WXDAFGJ844HXW7 -o microsoft.json

That ID retrieves this article's recorded run for its authorized owner; use the ID from your own command. Start with one company, confirm the returned fields, then approve the remaining URLs as a separate batch. The company-list automation walkthrough develops that repeated company enrichment workflow.

Why company fields matter before you build an agent

Company size, employees associated with LinkedIn and followers describe different things. Keep their original labels so the agent does not silently transform a page snapshot into an audited workforce claim.

Keep the workforce measures separate

The Microsoft response contained company_size equal to 10,001+ employees and employees_in_linkedin equal to 231115. One is a size label; the other is a numeric field describing employees in LinkedIn.

Do not rename both to headcount. If your account-scoring system requires a size bracket, use the bracket; if it tracks the returned LinkedIn employee measure over time, keep that number under its own name.

Treat followers as a separate signal

The Google response contained followers equal to 42682045. That value is a company-page audience measure in the retrieved record, not a count of customers, employees or reachable contacts.

For a research agent, the useful instruction is explicit: “Report follower count as follower count and explain no business outcome from it alone.” This keeps a large number from becoming an unsupported sales recommendation.

The broader LinkedIn data guide covers company, people and job-data jobs. Use that distinction when the next feature needs a different kind of record; adding a company field does not turn a company lookup into a people search.

The 3 ways to put company data into your workflow

Choose the route by how often you repeat the task and where the result needs to live. A one-time company check and an agent that refreshes a research table require different amounts of infrastructure.

Use manual review for a small decision

Read the company page yourself when you need context for a single meeting. A programmatic connection earns its place when you repeat the work, compare records or pass structured data to another application.

Manual review also gives you a place to resolve an ambiguous organization before automating it. A subsidiary and its parent company need separate input rows when your research question distinguishes them.

Use a shared tool connection for an agent

Monid fits a workflow in which the agent discovers and inspects tools before buying the calls it needs. The application still owns deduplication, field selection and the final table.

Keep the agent's tool permissions narrower than its research ambitions. Give it the company URL list, the allowed operation and a spending ceiling instead of asking it to collect everything it finds about each business.

Build directly when you need provider control

Choose a direct provider relationship when managing that connection is part of your product requirements. Keep your adapter responsible for credentials, response parsing and retention, regardless of where retrieval runs.

A script that reads pages itself also needs to handle changing page structure and extraction failures. That engineering belongs in the integration decision, rather than being hidden behind a three-line parsing example.

How do you read and export the results?

Read the result as a company snapshot plus an execution receipt. The receipt answers whether a call completed and what it billed; the snapshot contains the company fields to select for your research table.

Read the observed fields

The following values came back from the two company calls. Headquarters and industry are returned page fields, not independently verified statements about either organization.

Both inputs returned a company result, so this sample has two found companies and zero missing companies. It establishes the response shape for these inputs; it does not establish a coverage percentage for arbitrary company lists.

Export a deliberate selection

Ask your agent to write one CSV row per input company using the fields above, the requested company URL and the collection date. The source URL and date preserve where each row came from when the CSV leaves your application.

The returned website values were https://news.microsoft.com/ and https://goo.gle/3DLEokh. Preserve these as returned website values; do not automatically relabel either as a verified corporate root domain.

The response also contained nested material beyond the selected company columns. Export an allowlist of company fields instead of copying the whole response into a spreadsheet, a prompt or a customer-facing report.

For the next stage, repeat the same enrichment over your approved company URL list. Keep these field-selection rules when you move from two test companies to a longer account list.

Give missing values a useful meaning

Use an empty cell or explicit null for a field that is absent. Do not turn missing follower or employee data into zero: zero is a value, while missing means the record did not supply that value.

Store your own processing status separately from company data. Labels such as needs_review are application decisions, not fields returned by the provider, and the table should make that distinction visible.

How much does a company lookup cost?

The company-profile operation on Monid costs $0.008 per call, Verified 2026-11-11. Inspection declares PER_CALL, and each of our successful company requests billed one unit at that price.

Separate lookup price from batch size

The two calls together billed $0.008 each, Verified 2026-11-11. Multiply the current unit price by approved requests when setting your budget; repeated URLs and repeated refreshes are additional requests.

Discovery and inspection do not spend the balance. A paid lookup begins when you execute the operation, so reading a schema and reviewing a price belong before the loop.

Avoid paying twice for local processing

Save the completed response before formatting it. If CSV generation fails, repair the export from the saved JSON rather than repeating the paid company request.

Both test calls succeeded; this sample contains no failed or empty-response billing observation. Do not build a refund assumption into retry logic. Route unsuccessful requests to a bounded retry or review queue and retain their receipts.

Which lookup should you use for company research?

Use the tested company-profile lookup when your input is a LinkedIn company URL and your output is an organization research row. The same operation serves the following company tasks without requiring a different retrieval path.

Retrieve once and select multiple fields locally. Separate exports for industry and followers do not require separate company lookups when both fields are already in the saved result.

When should you use something else?

Use another route when your required outcome is outside company-profile retrieval. You are reading the Monid blog, and Monid routes paid provider calls; the right decision still starts with the job you need completed.

Choose your existing approved LinkedIn integration for account actions such as publishing or messaging. This company lookup supplies research data and does not perform those actions.

Choose TikHub directly when you want to manage the provider relationship yourself. Choose manual review when the decision concerns a single company and reading the page supplies the context you need.

The demonstrated operation accepts one company URL per call, and the slower test took 23.975 seconds of execution. Put it in a research job with saved results rather than treating it as an instant interactive lookup on every keystroke.

Conclusion

You now have a tested company lookup, a two-company result table and a clear separation between page metrics and business conclusions. Repeat the workflow by inspecting the current operation, approving a small list and saving each completed response before exporting through Monid. The opening estimate assumes 50 manual checks at two to three minutes each, less roughly 10 to 40 minutes for automated collection and review; the one-minute setup allowance is a planning estimate, not a measured onboarding result. Your next refresh can reuse the same field definitions and leave you with a comparable research table.

FAQ

Published November 2026.