Zevari vs Claude's own loops

"I'll just put my outreach on a Claude loop."

You run Claude Code or Codex, and you found /loop. It keeps a session alive, reruns your prompt on a timer, and even remembers what it did last pass. So the honest question is the one we keep getting: why not just loop your LinkedIn outbound and skip Zevari?

Here is the straight answer. A loop is a brilliant timer. Outbound is not a timer problem.

We are not here to talk you out of /loop. Zevari runs inside the loop - point the loop at Zevari and it drives a real campaign. The point of this page is narrower: a loop pointed at raw LinkedIn hits three walls fast, and if you have tried it, you have already met them. A loop gives you a clock. It does not give you hands on LinkedIn, durable state, or a safety model - and outbound dies without all three.

Connect to Claude CodeWe run it for you

Credit where due

What a loop is good at (and it really is good)

/loop was built for polling: check the deploy every five minutes, babysit a PR until CI goes green, retry until the work is done. It reruns your prompt inside one persistent session - same context window, same tool access, same MCP connections - so the agent remembers the last iteration and builds on it. Drop the interval and it self-paces. It is the right tool for short, self-terminating, in-session jobs.

That is exactly why it is the wrong tool for outbound. The strengths that make it great at babysitting a deploy - it lives in your session, it runs until a near-term stop condition, it has a hard 3-day ceiling - are the strengths that make it unable to hold a multi-week LinkedIn motion.

The three walls

Why a loop can't run LinkedIn outbound

Three walls, in the order you hit them.

01

The 3-day fuse

A loop runs for a maximum window and then auto-stops - no orphaned process, no runaway bill, by design. That is a feature for polling and a dealbreaker for outbound. A real campaign is a six-week, stateful motion: connect, wait, follow up on day 3, again on day 7, watch for the reply on day 12. A timer that expires Wednesday cannot run a sequence that lands its third touch next Tuesday. Re-arming the loop by hand every few days is just cron with extra steps - and the moment your session ends, so does the loop.

02

No hands on LinkedIn

Out of the box, Claude can research LinkedIn from the open web, but it cannot connect, message, comment, or send a request - there is no connection for it to act through. Looping a prompt that cannot touch LinkedIn just runs the same dead end on a schedule. The loop does not add hands; it adds repetition.

03

No safety model

The naive instinct is "loop: send 10 connections every hour." That is the precise pattern - unattended, unpaced, no ceiling, no human sign-off - that gets accounts restricted. A loop has no concept of a weekly connection ceiling, no human-like pacing, no burst caps, no staged approval. It does exactly what you said, as fast as the interval allows, until LinkedIn notices. The autonomous AI SDRs that ran unpaced churned most teams inside a year and at least one loud name got its accounts banned. A loop with no governor is how you join them.

A timer is not an engine

Zevari is the hosted engine the loop should be calling

Zevari is not a replacement for /loop. It is the durable, safe layer that makes looped outbound actually work - the hands, the hosted state past 3 days, and the safety governor a timer does not ship with. It is reachable over MCP for Claude Code and Codex, or our REST API from your own code.

The hands

Zevari is the hosted LinkedIn MCP: 60+ tools for search, signals, campaigns, inbox, and content. Because a loop shares your MCP connections, the same loop you would write can call Zevari to connect, message, comment, react, and post - the actions a bare loop has no way to perform.

Hosted state that outlives the window

Campaigns live on Zevari's infrastructure, not in a session that expires in 3 days or dies when you close your laptop. A campaign knows each target is on step 2 of 5 and advances it on schedule - next week, next month - whether or not your session is alive. The connection counter ticks against this week's ceiling between runs. Inbox Radar keeps watching replies after the loop has long stopped.

A safety governor, not a throttle you hope holds

Every write is staged for your approval and the platform enforces the limits for you: Free 40, Premium 80, Sales Navigator 150 per week, human-like pacing, burst caps, session-based connection with no browser cookies. You do not have to remember to pace your loop - you could not blow past the ceiling if you tried. A year of refinement, zero ban incidents.

Batched approval, not blocking mid-loop

Zevari keeps the human in the loop - that is why accounts stay safe - but it stages the whole queue at once: here are 15 connection requests and 6 messages, approve all from Slack in two minutes, instead of a loop iteration stalling on one confirmation at a time.

So the real comparison is not "Claude loop vs Zevari." It is "a 3-day timer with no LinkedIn hands and no brakes" vs "a hosted engine with all three." You bring the loop. We bring the layer it drives.

Side by side

/loop alone vs Zevari

Read it honestly: /loop is a fine way to schedule a prompt for a few days. It is not a way to run outbound.

Rerun a task on a timer
Claude /loop alone
Yes - that is what it is for
Zevari
Yes - hosted scheduling, no session required
Remember the last iteration
Claude /loop alone
Yes - within the session
Zevari
Yes - persistent campaign and contact history
Run past 3 days / laptop closed
Claude /loop alone
No - hard window, then auto-stops
Zevari
Yes - runs on Zevari's infrastructure
Actually send a LinkedIn message or connection
Claude /loop alone
No - no hands on LinkedIn
Zevari
Yes - 60+ native LinkedIn tools
Hold a multi-week, multi-step sequence
Claude /loop alone
No - state dies with the window
Zevari
Yes - targets advance on schedule for weeks
Enforce a weekly connection ceiling
Claude /loop alone
No concept of it
Zevari
Yes - 40 / 80 / 150, platform-enforced
Human-like pacing and burst caps
Claude /loop alone
No - fires as fast as the interval
Zevari
Yes - enforced server-side
Watch the inbox between runs
Claude /loop alone
Only while the session lives
Zevari
Yes - Inbox Radar runs in the background
Human approves before sending
Claude /loop alone
Possible, but blocks the loop per action
Zevari
Yes - whole queue staged, batch-approve in Slack
Connection method
Claude /loop alone
n/a - cannot connect
Zevari
Session-based, no browser cookies
Net result for outbound
Claude /loop alone
A timer that expires before the campaign does
Zevari
An engine that runs the pipeline while you approve

The right column is what you point the loop at to make the left column's ambition survive past Wednesday - and survive LinkedIn.

The governor a loop does not have

Every write staged for approval, session-based connection with no browser cookies, weekly ceilings of 40 / 80 / 150 enforced server-side, plus human-like pacing and burst caps. A year of refinement, zero ban incidents.

Read the safety model

The question, answered

"Can a Claude loop run my LinkedIn outreach?"

Short answer: not on its own, and not safely. A loop can rerun a prompt, but the prompt still has no native way to connect to LinkedIn, the loop expires in 3 days, and nothing in it paces your sends or enforces a ceiling - so an unaided outbound loop is both blind and reckless.

Zevari is what the loop calls. Reach it over MCP for Claude Code and Codex, or our REST API from your own code, and your loop - or a routine, or a single prompt - can drive real campaigns: every send staged for approval, weekly ceilings enforced, state held on our infrastructure long after the loop window closes. You are not getting a different AI. You are giving the AI you already loop a set of hands and a governor.

FAQ

Questions real buyers ask

Can I just put my LinkedIn outreach on a Claude loop?

Not for real outbound. /loop reruns a prompt inside one session for a maximum of about three days, then stops - so it cannot hold a multi-week campaign. It also has no native way to connect to LinkedIn and no concept of a weekly ceiling or pacing, so an unattended outbound loop is both unable to send and unsafe when it can. Zevari adds the LinkedIn hands, the hosted state that outlives the window, and the enforced safety limits - then batches approvals so you sign off on a whole queue at once. Zevari is reachable over MCP for Claude Code and Codex, or our REST API from your own code.

What is the difference between a loop and a routine for this?

A loop runs inside your live session and remembers across iterations, but it is ephemeral - 3-day max, dies when the session ends. A routine runs on the cloud with your laptop closed, but each run is more stateless. Both schedule an agent that still has no LinkedIn hands and no ban-safety. Zevari is the layer either one calls to get hosted state and safe sending, reachable over MCP for Claude Code and Codex, or our REST API.

Why can't I just enforce my own pacing in the loop prompt?

You can ask, but nothing makes it true. A loop will fire on its interval regardless, and a hand-rolled wait in a prompt is not a platform-enforced ceiling, human-like pacing, or a burst cap. Zevari enforces 40 / 80 / 150 weekly limits and pacing server-side - you could not exceed them if you tried. That is the difference between hoping you are safe and being unable to be unsafe.

Does Zevari replace /loop?

No - Zevari is the hosted engine the loop calls, and a loop is a perfectly good way to trigger Zevari for a short, supervised burst. What Zevari replaces is the missing parts: the hands on LinkedIn, the campaign state that survives past 3 days, and the safety governor. Loop us if you like; we hold the pipeline either way.

Do I still need Claude Code or Codex if I use Zevari?

Zevari is reachable over MCP for Claude Code and Codex, or our REST API from your own code. It gives your agent hands on LinkedIn and a place to keep state. If you do not run Claude Code or Codex at all, our managed tier runs the whole thing for you and you approve every send from Slack.

Is unattended, looped LinkedIn outbound even a good idea?

Only with a governor. Fully autonomous, unpaced sending is what got the loud AI SDRs banned and churning inside a year. The operator consensus is that human-in-the-loop beats autonomy on LinkedIn. Zevari's posture - every write staged for approval, no browser cookies, enforced ceilings - is exactly that. A year of refinement, zero ban incidents.

You brought the loop. We brought the engine.

Reach Zevari over MCP for Claude Code and Codex, or our REST API from your own code. Two ways to run it - pick your lane.

Connect to Claude Code

You run Claude Code or Codex and you reached for /loop. It got you a timer and stalled at "it cannot touch LinkedIn and it expires Wednesday." Zevari is the missing layer: outbound-as-code, hosted state past the window, enforced ceilings, batched approval. Live in 60 seconds.

Connect to Claude Code

We run it for you

You would rather not touch loops, prompts, or a terminal - you want pipeline. We run the same approval-gated engine and you approve every send from Slack in minutes a day.

We run it for you

Zevari - the LinkedIn execution layer for Claude. Your loop expires; your pipeline does not.