Apex

your LinkedIn on autopilot — a loadout for AI agents

Not a raw API. Outcome commands that let an agent run your whole inbound loop, bounded by values you set and account health you can’t fake.

The apex CLI on this page is a preview — the commands aren’t shipped yet (anything tagged soon is still in the works).
The capabilities behind it are live today via MCP and the REST API: Discover & Signals (Apex), post scheduling & automations (Pro), and company-page actions (Pro+).

What a loadout is

A loadout is a self-improving loop that can see, decide, and act.

See

Read signals, leads, engagement, and what’s converting — its eyes on the field.

Decide

Score who’s worth acting on and what to say, using fit, intent, and live performance.

Act

Publish, comment, deliver, follow up — at your pace, never beyond your limits.

Overview

One agent, the whole loop

Apex closes the loop most tools leave open. An engagement tool acts but can’t see your leads. A content tool decides but can’t see who engaged. An automation runs rules but holds no strategy. Apex holds every context at once — so it does the job of the whole stack, with shared memory, and gets better each cycle. It’s driven by the apex command. Every command maps to a dashboard action you already trust, bounded by the active safety mode, your daily limits, and account health.

Three layers, one surface

apex — the Apex CLI, the command surface (this doc).

MCP — the same actions as tools for in-chat agents.

REST API — the primitives everything is built on.

Why a loadout wins

Precise, not unlimited.

Shared context

One brain sees leads, signals, and performance at once — the thing point tools structurally can’t copy.

Self-improving

Each cycle reads the result of the last. The loop compounds; the strategy sharpens on its own.

Governed velocity

Per-account budgets, cooldowns, human-jitter timing. Built to keep your account alive.

Observable by design

Audit log, scopes, kill switch. You wake up to a report, not a surprise.

Surfaces

Everything Apex can do, grouped by what you’re acting on. Pick a surface, then a verb.

Personal

apex personal

Act as your own LinkedIn profile — comment, react, reply, DM, and manage your inbox.

commentreactsoonreplysoondmsooninboxsoon

Company

apex company

Act as a Company Page you administer — list your pages, comment, react, and invite leads to follow.

listcommentreactfollow-invite

Posts & automations

apex posts

Draft and schedule posts (as you or a Company Page), add a first comment, and wire comment-to-DM automations that activate on publish.

schedulefirst-commentautomateeditactivate

Pages & links

apex pages

Capture / quiz landing pages and tracked lead-magnet links, with click and email analytics.

createlistlinksanalytics

Leads & signals

apex signals

Intent signals, captured leads, and hot leads — with heat and ICP fit scores.

signalsleadshotscore

Discover

apex discover

Lead Magnet Radar — surface the lead-magnet plays converting in your niche (live via MCP & API) — plus an AI post-writing skill that drafts posts in your voice from what’s working.

radarwriteengineersoonplansoon

The Loop

Each stage of the loop draws on the surfaces it needs — same vocabulary, end to end.

Create

Draft the assets the loop runs on.

Posts & automationsPages & linksDiscover

Publish

Ship content as you or a Company Page.

PersonalCompanyPosts & automations

Capture

Turn engagement into signals and leads.

Leads & signals

Qualify

Rank and score who to act on.

Leads & signals

Follow up

Reply and DM the right people.

PersonalCompany

Protect wraps the whole loop — limits, cooldowns, suppressions, and audit apply to every stage.

Install & Auth

The apex CLI is rolling out. Today the same capabilities are live via MCP (install @leadshark/mcp-server or paste the server URL into Claude.ai) and the REST API. The CLI steps below are the design preview — ask us for early access.

Requirements

• Node.js 18+
• LeadShark Apex subscription (full agentic LinkedIn is the Apex tier)
• API key from Dashboard Settings, or sign in over OAuth

Install

npm install -g @leadshark/apex

Authenticate

Interactive (opens a browser, OAuth device flow):

apex login

Headless (CI / agents) — set an API key in the environment:

export LEADSHARK_API_KEY="sk_live_..."
apex whoami

$ apex whoami
account   ruzgar@leadshark.io
tier      Apex
mode      draft         (no live actions until you switch)
linkedin  connected · healthy
limits    DM 42/80 today · cooldown clear

The CLI never holds standing write access. It uses your existing LeadShark permissions and the active safety mode — calls fail closed if your account is not Apex or LinkedIn is disconnected.

Safety ModesSoon

The values are the throttle. A loop that can see, decide, and act with no values is the thing that gets an account banned. Apex’s power isn’t volume — it’s judgment.

The mode is the single lever that bounds what the agent may do. It applies to every command. Start indraft and escalate deliberately.

Mode	What the agent may do
`read-only`	Read and list only. No writes of any kind. Safe for analysis and planning.
`draft`	Create assets, posts, and automations as drafts. Nothing goes live or sends.
`queue`	Stage actions (posts, replies, DMs) for human approval. Nothing leaves until approved.
`live`	Act automatically within your daily limits, cooldowns, and suppressions.
`admin`	Account-level controls: set limits, suppress, emergency stop. Reserved for owners.

$ apex mode queue
mode changed: draft -> queue
queued actions now require: apex followup approve <id>

A command that requires a higher mode than the current one is refused with a non-zero exit code and a hint — it never silently escalates. Per-command overrides use--mode, which still cannot exceed your account ceiling. High-risk reach like InMail stays gated here until the layer is enforced, by design.

Commands

The loop, as a surface. Each command does one thing, prints human output by default, and supports--format json for agents. Grouped by the loop stage it serves.

Every command carries its values (what it won’t do) and know-how (how this is won).

SEEread the field

Capture

Turn the field into structured records.

DECIDErank who to act on

Qualify

Rank who’s worth acting on.

ACTmake and ship

Create & publish

Make and ship the assets the loop runs on.

Publish

Ship content as yourself or a Company Page, plus comments, replies, reactions, and InMail.

Follow up

Reply & DM the right people — automations, drafts, queue, and inbox.

VALUESthe throttle on every stage

Protect

The values that wrap every stage.

Recipes

Composed flows an agent runs end to end. Modes keep each one as safe as you want it.

Plan a lead-magnet launch (human-approved)

apex mode queue
apex create magnet --name "Cold Guide" --destination https://example.com/guide.pdf --capture-email
apex create post --brief "Launch the cold guide" --cta "Comment guide"
apex publish post --id <post_id> --at "Mon 9am EST"
apex followup queue          # review what will go out
apex followup approve --all  # release it

Work the inbound after a post pops off (live mode, bounded)

apex limits set --dm 80
apex mode live
apex capture watch --events new_comment &
apex qualify --icp icp_founders --update
apex opportunities list --min-heat 75
apex followup draft --opp <id> --template "Hey {{firstName}}, here is the guide: {{link}}"

Daily standup for an agent (read-only)

apex mode read-only
apex health
apex opportunities list --min-heat 70 --format json
apex inbox --state awaiting
apex audit --since 1d

Guardrails

These apply to every command, in every mode — they are not optional flags.

Daily limits

Every send checks your DM/connection budget. Over budget = queued, never dropped silently.

Cooldowns

On LinkedIn rate-limits the CLI backs off automatically and reports the cooldown window.

Suppressions

Do-not-contact entries are honored across all stages, including live mode.

Account health

If LinkedIn is disconnected or flagged, write commands fail closed with a clear reason.

Audit log

Every action is recorded with timestamp, mode, and result. apex audit tails it.

Dashboard parity

Every command maps to an action you can see and undo in the dashboard.

Scripting & JSON

Built for agents and pipelines: structured output, predictable exit codes, and a global--mode override.

JSON output

$ apex opportunities list --min-heat 80 --format json
[
  { "id": "opp_03", "name": "Omar K.", "heat": 91, "icp_fit": "strong" }
]

Exit codes

`0`	Success
`2`	Refused by safety mode (escalation required)
`3`	Blocked by guardrail (limit, cooldown, suppression, or health)
`4`	Not authorized (tier or LinkedIn not connected)

Prefer not to shell out? The same surface is available as MCP tools and the REST API. The CLI is the outcome-shaped layer on top.

Questions or early access?

Apex is our flagship agent — your LinkedIn on autopilot. Reach out for onboarding.

Contact