Free — 7 questions, no signup, no model call

Build a LinkedIn AI stack, layer by layer

A LinkedIn AI stack has 4 layers. One of them is a subscription; the one that decides what comes out is free. The seat you pay LinkedIn for, the written method that costs nothing, the data layer and the agent layer. Answer 7 questions and this page assembles one from 9 LinkedIn plans, 42 third-party tools and 13 MCP servers, with the figure, the source and the account exposure beside every part. Of the 4725 distinct answers it can be given, 1810 come back saying buy nothing.

✓ Every figure read from the seller's own page, and dated✓ 14 of 42 tools are ones it will name — the rest put your account behind them✓ Same answers, same stack: a decision tree, not a model call✓ No affiliate links, no sponsored placements

Answer the 7 questions →What LinkedIn itself costs

True whatever you answer

What a LinkedIn AI stack is made of

Four layers, in the order that decides the outcome rather than the order of the invoice. A stack page that led with tools would be saying that buying is the first move.

1
The seat
Sold by: LinkedIn
The subscription LinkedIn sells you. It sets the InMail allowance, the search depth and the filters, and nothing above it can lift a ceiling it leaves in place.
7 of LinkedIn’s 9 plans publish a price, from $39.99 to $170 per seat per month. 2 are quoted through a representative, and this page prints no figure for those. Read 12 September 2026.
What every LinkedIn tier costs
2
The method
Sold by: You
The workflow, the prompts and the context files an assistant reads before it writes anything. It is the layer that decides what the output is worth, and the one people skip.
Free. 28 prompts, 15 context files, 7 agent packs and 4 workflows, all on this site. This layer is in every stack the tool below builds, including the ones that recommend buying nothing.
The AI toolkit
3
The data layer
Sold by: A vendor
A third-party product that finds, enriches or watches — lists, signals, company records. What decides whether it belongs in a stack is not what it does but what it puts behind it.
42 tools in the directory, read 20 September 2026. 14 of them reach LinkedIn with no account of yours behind them, and the tool below will name one only from that set; 18 of the 42 act under your name instead.
The tool directory
4
The agent layer
Sold by: A repository
An MCP server, which is what lets an assistant fetch something itself rather than wait for you to paste it. It is the last layer worth adding and the easiest to add wrongly.
13 servers in the directory, read 20 September 2026, none of them LinkedIn’s. 4 need no LinkedIn account. Not one publishes a monthly seat price, so nothing in this layer is ever added into a total.
The MCP directory

And 4 ceilings no layer moves

Every one of these applies on a free account and on every paid tier, including the ones LinkedIn will not price. A stack is planned inside them, or it is planned against a number that does not exist.

Connection invitations~100 per week
Applies on every plan including Recruiter Corporate. LinkedIn adjusts it per account based on acceptance rate, so a low-acceptance account gets less, never more.
InMail credit expiry90 days
Unused credits expire. They accrue to at most three times the monthly allowance, and a credit is returned if the recipient replies within 90 days.
Commercial use limitMonthly search cap
Free accounts hit a monthly ceiling on profile searches. Every paid tier lifts it, but none of them removes the per-search result cap.
Profiles per search1,000 / 2,500
Regular LinkedIn search stops at 1,000 results; Sales Navigator stops at 2,500, paginated 25 at a time. Both are per search, not per day.

Read 12 September 2026. The full table, by plan, is on LinkedIn limits by plan.

Read before you assume there is something to buy

What it says when the answer is not a product

Run over all 4725 answers it can be given, the engine ends in a stack with something to buy in 1340 of them. The other 3385 end in one of these. They are the reason the page exists, not error states in it.

Buy nothing — 1810 of 4725

A free account, everything written and sent by hand, no budget. The output is the account you already have with its ceilings named, plus the workflow, the pack, the prompts and the context files. The seat the job would otherwise want is listed as out of reach with its real figure.

No sanctioned route — 1575 of 4725

Ask for something that sends or posts under your name and the page leads with the fact that the capability exists and the permission does not. It prices each route by class, quotes each class's own ceiling, and names no vendor.

Out of reach — 2513 of 4725

The capability sits behind a tier the budget does not reach. It is named with the real figure and the shortfall, and the stack falls back to the free account rather than to a cheaper plan that does not do it. Some of it has no published price at all, and then the total is a floor.

Nothing to name — 2250 of 4725

For several jobs the tool directory holds nothing this page will name, and says so in the directory's own words rather than stretching a prospecting tool to fit. Two of 45 researched categories came back empty and are recorded as gaps.

A recommender that always recommends a purchase is an advertisement. This domain sold LinkedIn account rental until 10 September 2026 and sells nothing of the kind now, which is a reason to be further from that than average rather than closer.

The form, written out

The 7 questions, and what every answer moves

The same options the tool shows, with the consequence of each one spelled out — readable without a form in the way, and the reason the tool above prints one line under each row instead of five.

What do you use LinkedIn for?

Picks the workflow, the pack, the prompts and the LinkedIn product family. It also decides whether the tool directory holds anything at all for you — for several of these it holds nothing, and the output says so rather than stretching a prospecting tool to fit.

Sales and outbound Sales Navigator becomes the seat question, and the tools that reach LinkedIn without an account of yours become available.
Recruiting Recruiter becomes the seat question, and the output says how many of the recruiting tools put your own account behind them.
Posting and content No seat is recommended and no tool is named — the directory declares a gap here, and LinkedIn withdrew its own post writing tool.
Finding a job Premium Career short-circuits the seat question, because the features a job search needs exist on that tier and on nothing dearer.
Networking and being found The answer is usually the free account: the one ceiling paying lifts for this job is one most people never reach.

What LinkedIn seat do you hold today?

Decides whether the seat layer is a purchase, an upgrade or a line saying you already hold what this job needs. It also sets the InMail and search ceilings the rest of the stack has to plan inside.

Free account The seat layer becomes a purchase decision, and the commercial use limit applies.
Premium Career The job-search features are already yours; for selling, it carries the smallest InMail allowance LinkedIn sells.
Premium Business The commercial use limit is already lifted, so no upgrade is recommended for browsing volume.
Sales Navigator (Core or above) Lead lists and the advanced search are already yours; the question becomes whether LinkedIn’s own Sales Navigator AI, which starts one tier up, is worth the difference.
Recruiter Lite Sourcing is covered inside your own network; the output prices what reaching past it costs.

Alone, or with others on the same accounts?

The only question that moves the seat up a tier: shared lists and TeamLink are the Core/Advanced line in Sales Navigator, and shared pipelines are the Lite/Corporate line in Recruiter.

Alone The team tier is not recommended, and the output prices what it would have cost.
With a team The team tier is recommended, and in Recruiter that tier carries no published price at all.

How much of this should run without you?

The most consequential answer here. It decides whether the stack is a prompt, a research layer, or a request for something no sanctioned route provides — and it stands in for a volume question, because someone who writes and sends by hand is not working at a volume any tool changes.

I write and send everything No tool and no server is recommended. The output is the platform ceilings and the hand-written prompt for the job.
Research and drafts, but I send The data and agent layers open up, restricted to routes that put no LinkedIn account of yours behind them.
I want it to send The output leads with what has no sanctioned route, prices each unsanctioned route by exposure, and recommends none of them.

What can this cost per month, in total?

Applied to every published figure in the stack. Where the capability you asked for sits behind a tier your budget does not reach, the output says so with the real figure instead of handing you a cheaper thing that does not do it.

Nothing Only free tiers and free resources are recommended; everything else moves into the out-of-reach list with whatever figure its seller publishes.
Up to $50 Reaches the cheapest LinkedIn seat and the cheapest of the tools this page will name, and nothing else.
Up to $150 Reaches the single-seat tiers at the cheaper end. The dearer seats and the enrichment layer stay out of reach, and the output states the gap in dollars.
Up to $500 Reaches every LinkedIn seat that carries a published price, and the paid tools this page will name.
More than $500 Nothing is excluded on price. The plans LinkedIn will not price are still marked unpriceable rather than affordable.

How much setup are you willing to do?

Measured against each record’s own setupDifficulty, and against whether a server runs on someone else’s machine or on yours. It is also the gate on the whole MCP layer: a config file is the floor, and pretending otherwise produces a recommendation nobody installs.

None — it has to work in a browser No MCP server is recommended, and tools the registry marks hard to set up are excluded.
I can edit a config file Hosted MCP servers become available. Ones you would have to run yourself do not.
I run things on my own machine Self-hosted servers and the hard-to-set-up tools become available, which is where the free options are.

Which assistant do you use?

Matched against the client list each MCP server publishes. It decides which server, if any, you can actually connect — not which is better.

Claude Every server that needs no LinkedIn account names a Claude client, so the choice is made on hosting and cost instead.
ChatGPT Narrows the agent layer to the servers that name ChatGPT as a client.
Gemini Narrows the agent layer to the servers that name a Gemini client.
None yet No agent layer is recommended. The prompts work in any of them, so nothing is lost that can be lost later.

And the 5 it does not ask

A question that cannot move the output costs the reader an answer and makes the result read as a quiz. Each of these was considered and dropped, with the derivation rather than a shrug.

Do you hold Sales Navigator? Do you hold Recruiter?: Both are answered by the seat question. They are not independent of it: what decides the output is the highest seat held, so a second and third question whose answers the first already implies would add two steps and no branch.
Company size: Nothing in the output keys on headcount. What people mean by it is either the team question — does anyone else work these accounts — or the budget question, and both are asked directly. LinkedIn’s own tiers are per seat, with no published volume break at any size.
Which CRM do you use?: It cannot change which component is chosen. Of the tools this engine will name, every one that integrates HubSpot integrates Salesforce and the other way round, so neither answer excludes anything; the one that integrates neither publishes no CRM connector at all and is named for watching public posts rather than for keeping records. Derived in COUNTS.crmCoverage and asserted at module load, so the claim fails the build if it stops being true.
Team size as a number: The seat tiers change at one boundary, not on a slope: TeamLink and shared lists in Sales Navigator, shared pipelines in Recruiter. A number would imply the recommendation moves at three seats and again at ten, which it does not.
How much LinkedIn experience you have: It changes how something is explained, not what is recommended, and the resources it would point at are the same ones. The technical question is asked instead, because that one genuinely closes a layer.

About this tool

What it is and is not

Does this call an AI model?

No. It is a decision tree over the registries this site already publishes, so the same answers always produce the same stack — which is what makes a recommendation checkable rather than persuasive, and what lets a coverage script in the repository run the engine over all 4725 distinct answers it can be given rather than over a few someone thought of. The registries themselves are resolved when the module loads, so a stack that named a prompt, a plan or a vendor tier that had been renamed would fail the build rather than reach a reader.

Will it ever recommend something that automates LinkedIn for me?

No, and the reason is one rule rather than a policy paragraph: it will name a third-party tool only where the route that tool takes puts no LinkedIn account of yours behind it. That is 14 of the 42 records in the directory, and 4 of the 13 servers. Answer that you want something to send on your behalf and the page leads with the fact that no sanctioned route reaches it, prices every unsanctioned route by class, and recommends none of them. The directories list them all; a recommender may not pick one.

Where do the prices come from?

LinkedIn figures come from LinkedIn’s own US list prices, read 12 September 2026, through the same data layer as every pricing page on this site. Vendor figures are the vendor’s own tier, quoted verbatim as it read on 20 September 2026. Where a seller publishes no figure, the page prints “Quote only” and no number — 2 of LinkedIn’s 9 plans are in that position, and the total above them is stated as a floor rather than as a price.

Why does answering a question change the address bar?

So the answer is a link. A recommendation that exists only in the tab it was made in cannot be sent to whoever approves the spend. It also means the recommendation is built on the server, which is why this page works with JavaScript off and why a search engine sees a complete stack rather than an empty form.

It did not ask which CRM I use. Why not?

It cannot change which component is chosen. Of the tools this engine will name, every one that integrates HubSpot integrates Salesforce and the other way round, so neither answer excludes anything; the one that integrates neither publishes no CRM connector at all and is named for watching public posts rather than for keeping records. Derived in COUNTS.crmCoverage and asserted at module load, so the claim fails the build if it stops being true.

Where to next

The method layer is free, and it is the one that decides the output

28 prompts, 15 context files and 4 workflows, written and edited rather than generated. Every stack above starts with them, including the ones that recommend buying nothing.

The AI toolkitThe whole library, with what each piece is for.All 28 promptsDecision logic, anti-patterns and a worked example in each.

Build a LinkedIn AI stack, layer by layer

What a LinkedIn AI stack is made of

The seat

The method

The data layer

The agent layer

And 4 ceilings no layer moves

Answer 7 questions

Nothing here is worth buying. Here is what to do instead

How it got there

The seat

The account you already have

The method

From ICP to booked meeting

LinkedIn Sales Agent

ICP Builder

Sales Navigator Search Builder

LinkedIn Outreach Personalizer

Follow-up Sequence Builder

ICP.md

LINKEDIN-LIMITS.md

OUTREACH.md

Named, and out of reach

LinkedIn's own AI for this job

What it says when the answer is not a product

The 7 questions, and what every answer moves

And the 5 it does not ask

What it is and is not

The method layer is free, and it is the one that decides the output

Build a LinkedIn AI stack, layer by layer

What a LinkedIn AI stack is made of

The seat

The method

The data layer

The agent layer

And 4 ceilings no layer moves

Answer 7 questions

Nothing here is worth buying. Here is what to do instead

How it got there

The seat

The method

Named, and out of reach

LinkedIn's own AI for this job

What it says when the answer is not a product

The 7 questions, and what every answer moves

And the 5 it does not ask

What it is and is not

Continue researching

The method layer is free, and it is the one that decides the output