Skip to content

cli.md

Latest commit

 

History

History
120 lines (92 loc) · 5.2 KB

cli.md

File metadata and controls

120 lines (92 loc) · 5.2 KB
Raw

lit — the CLI

Ships in the same npm package as the MCP server, so npx linkedin-toolkit-mcp already installed it. Same engine, same caps, same approval queue as everything else.

npm i -g linkedin-toolkit-mcp   # or just use npx
lit status

Server

lit serve --http                       # Streamable HTTP MCP + /actions + /openapi.json on 127.0.0.1:47830
lit serve --http --port 8080           # different port
lit serve --http --fake                # fake extension client: no LinkedIn account needed, real envelopes

GET /health for a liveness check, GET /openapi.json for the generated OpenAPI 3.1 document. Bearer token is token in ~/.linkedin-toolkit/config.json, or run lit config get token --reveal.

Diagnose

lit endpoints check                    # self-test every LinkedIn endpoint the extension uses
lit endpoints check --post <post-url>  # also check the reactions endpoint
lit endpoints check --json             # the raw status.get { verify: true } envelope

Run this first when anything returns LINKEDIN_ERROR. One read-only call per endpoint, reported as ok, failed, unverified (never checked against the current LinkedIn client) or skipped (nothing on this account to check it against), each with the client version the endpoint table was captured against. Exits 2 if any endpoint failed, so CI can watch for a LinkedIn release breaking a query id: a failed row means LinkedIn moved, not that the toolkit is broken — see voyager-endpoints.md for how to recapture. The pass costs one search result and one profile visit against the daily caps, metered exactly as the real actions are.

Read

lit status                                                     # quota, mode, queue, campaigns
lit search "CTO fintech London" --source salesnav --count 100 --csv out.csv
lit profile https://www.linkedin.com/in/... --full --json
lit engagers <post-url> --list "Post engagers 8 Sep"
lit inbox --since 24h --sentiment

Write

Everything here queues in Copilot mode. --dry-run on any of them prints what would be sent and touches nothing.

lit invite <profile-url> --note "..."
lit message <profile-url> --body "..."
lit campaign create --from sequences/warm-connect.json --list "Data leads"
lit campaign pause <campaignId>
lit queue list                                                # pending by default
lit queue list --status sent|failed|approved|pending|rejected
lit queue approve <id> [<id>...]                              # marks them; sending follows
lit queue reject  <id> [<id>...]

lit queue approve returns as soon as the items are marked. The extension then sends them one at a time at human pace — seconds to minutes, depending on your delay settings — so check the outcome with lit queue list --status sent and --status failed, which shows why each one failed. A note longer than LinkedIn's 200 characters is refused at approval time, and nothing is approved.

Research Pack

lit research input.csv --out ./packs                  # resolve, capture, signals, dossier per row
lit research input.csv --out ./packs --resolve-only   # dry-run the matching before spending quota

Writes pack.md and pack.json per row plus an enriched output.csv, and creates a list. A large CSV is spread over days by the engine; the ETA is printed up front. Use the linkedin-research-pack skill to add the public-web layer with your agent's own search tool.

Local data

lit sql "select company, count(*) from profiles group by 1 order by 2 desc limit 20"
lit export --table profiles --csv
lit sync                                              # pull everything changed into SQLite

SELECT only, over ~/.linkedin-toolkit/toolkit.db. No network, no quota.

Config

lit config get                                        # the whole config, token redacted
lit config get webhookUrl                             # one key
lit config get token --reveal                         # print the bridge token in full
lit config set webhookUrl https://your-n8n/webhook/linkedin-events
lit config set dailyInviteCap 20
lit token rotate                                      # new bridge token; re-pair the popup after

Hard ceilings are clamped regardless of what you pass: 100 invites, 150 messages, 500 profile visits, 1,000 search results per day. Autopilot cannot be set from here — it is a toggle in the extension popup and only a human can flip it.

lit token rotate invalidates the current bridge token immediately, so anything holding it — the extension, a running client, a tunnel you handed to a remote agent — stops working until you re-pair. That is the point of it.

Environment

Two variables, both optional. The Node and Python clients read them; so does lit when it is talking to an already-running server.

Variable Default What
LINKEDIN_TOOLKIT_URL http://127.0.0.1:47830 Where the HTTP surface is
LINKEDIN_TOOLKIT_TOKEN — The bridge token, otherwise read from token in ~/.linkedin-toolkit/config.json

Full action reference: actions.md. Shell examples: ../examples/cli/.