linkedin-cli

linkedin

A command line for public LinkedIn data. One binary that turns member profiles, company pages, job postings, and the guest job search into rich, structured records as a list, table, Markdown, JSON, JSONL, CSV, TSV, or plain URLs.

linkedin jobs "golang backend" --location Remote -n 5

JOB_ID      TITLE                       COMPANY    LOCATION                   POSTED
4391940951  Backend Engineer (Go)       Xsolla     Montreal, Quebec, Canada   2025-12-01
4388210773  Senior Go Engineer          Datadog    Remote                     2025-11-29
...

Full documentation: linkedin-cli.tamnd.com.

linkedin is an independent, open-source tool. It is not affiliated with, endorsed by, or sponsored by LinkedIn or Microsoft. It reads only public pages, with no account and no API key, at a polite default rate.

Why

LinkedIn's public data sits behind server-rendered pages and a handful of guest endpoints, with no open API to read it. Pulling anything structured out means hand-rolling a scraper, guessing at selectors, and redoing the work whenever a page changes. linkedin puts the public surface behind one tool with sensible defaults, real output formats, and pipelines that compose. It reads each page JSON-LD first and falls back to HTML selectors, so a profile or a company comes back as a complete record, not a bag of strings.

It speaks to www.linkedin.com over plain HTTPS. The binary is pure Go with no runtime dependencies.

What works anonymously, and what is walled

LinkedIn serves some surfaces to anonymous visitors and gates the rest behind a sign-in wall. linkedin reads what is public and reports clearly when a page is walled.

One detail makes the HTML surfaces work: an anonymous request that sends a LinkedIn self-Referer is treated as a scraper and answered with HTTP 999, so linkedin sends no referer at all and the profile and company pages come back normally. The dedicated activity and /posts/ subpages still redirect to the sign-in wall, so posts are read from the main page's JSON-LD instead.

When a page is gated, linkedin exits with code 4 and you can lend a signed-in session with --cookies (a Netscape cookies.txt jar exported from your browser).

Install

go install github.com/tamnd/linkedin-cli/cmd/linkedin@latest

Or grab a prebuilt binary from the releases page, install a Linux package (deb, rpm, apk), or pull the container image:

docker run --rm ghcr.io/tamnd/linkedin jobs "golang" -n 5

Homebrew and Scoop:

brew install --cask tamnd/tap/linkedin
scoop install linkedin

Build from source:

git clone https://github.com/tamnd/linkedin-cli
cd linkedin-cli
make build      # produces ./bin/linkedin

Quick start

linkedin profile williamhgates              # a member profile as a record
linkedin profile williamhgates --posts      # the member's recent public posts
linkedin profile williamhgates --articles   # the member's long-form articles
linkedin company microsoft                  # a company page as a record
linkedin company microsoft --posts          # the company's recent public posts
linkedin company microsoft --locations      # the company's offices
linkedin company microsoft --affiliated     # the company's related pages
linkedin jobs "golang backend" --location Remote   # job stubs from the guest search
linkedin job 4391940951                     # a full job posting
linkedin jobs "data engineer" --hydrate -n 20 --save  # full jobs, into the store

How it works

linkedin reads the same pages a logged-out browser sees and normalizes each one into a struct, with explicit empty, zero, or [] for fields that are genuinely absent. Most public pages carry a JSON-LD block; linkedin reads that first and uses CSS selectors to fill in the rest. Responses are cached on disk (content-addressed and gzipped) so a repeat call is instant and does not hit the network.

• profile reads the Person JSON-LD a public profile ships: name, headline, location, follower count, current roles, schools, and board or group memberships. With --posts it emits the member's recent posts and with --articles their long-form articles, both carried in the same JSON-LD graph.
• company reads the Organization JSON-LD and the company about panel: name, description, website, follower count, headquarters address, employee count, industry, size band, type, founding year, specialties, logo, and the funding round count with a Crunchbase link to the latest round. With --posts it collects the DiscussionForumPosting nodes the page carries, with --locations the full office list, and with --affiliated the related affiliated and showcase pages.
• jobs reads the anonymous guest job-search endpoint, paginating in pages of 25 until -n results are gathered or the endpoint runs dry. With --hydrate it follows each stub to the full job record.
• job reads the guest job-detail fragment: title, company, company logo, location, applicant count, posting date, the full description, and the criteria block (seniority, employment type, function, industries).

When a page is behind the sign-in wall or returns LinkedIn's bot block (HTTP 999), linkedin exits cleanly with code 4.

Commands

Output

Output is a readable list view on a terminal and JSONL when piped, so it drops straight into a pipeline. Use -o table for the bordered grid and -o markdown for a paste-ready table. Pick any format explicitly with -o:

linkedin company microsoft -o json           # pretty JSON array
linkedin jobs "golang" -o jsonl              # one JSON object per line
linkedin profile williamhgates -o csv        # CSV with a header row
linkedin jobs "golang" -o url                 # just the job URLs
linkedin jobs "golang" -o markdown            # a GitHub-flavored pipe table
linkedin job 4391940951 --fields title,company,location -o tsv
linkedin profile williamhgates --template '{{.Name}} has {{.Followers}} followers'

Choose columns with --fields, drop the header with --no-header, and apply a Go text/template per record with --template.

Storing records

Records can be upserted into a local SQLite store, keyed by kind and id, so you can build a dataset over many calls and query it back:

linkedin jobs "data engineer" --hydrate -n 50 --save   # fetch and store
linkedin company microsoft --save                       # store one company
linkedin db count                                       # how many records, by kind
linkedin db query --kind job -n 10                      # read them back

The fetcher is polite by default (a 2s spacing between requests) and you can tune it with --rate.

Exit codes

Configuration

State lives under $XDG_DATA_HOME/linkedin (or ~/.local/share/linkedin), overridable with --data-dir or LINKEDIN_DATA_DIR. The page cache and the SQLite store both sit there. Politeness and networking knobs (--rate, --timeout, --retries, --cache-ttl, --no-cache, --refresh, --cookies) are global flags on every command. Run linkedin info to see the resolved paths and linkedin <command> --help for the full surface.

Development

make build      # build ./bin/linkedin
make test       # go test ./...
make vet        # go vet ./...
make fmt        # gofmt -s -w .

License

Apache-2.0.