Search for the LinkedIn API and you find two different worlds. The official one is a set of partner programs that most developers will never be approved for. The practical one is an ecosystem of third-party infrastructure that products actually build on. This guide covers both, and what to use depending on what you are building.
Does LinkedIn have an official API?
Yes, but not the one most people are looking for. LinkedIn's developer platform is organized around business lines, each with its own partner programs: Consumer, Compliance, Learning, Marketing, Sales, and Talent. The Marketing API Program covers advertising, community management, and lead sync; the Sales Navigator Application Platform (SNAP) covers CRM integrations; Talent Solutions covers recruiting. LinkedIn's own documentation states that most permissions and partner programs require explicit approval, with only a small set of Open Permissions available to every developer, and the granted scopes stay narrow.
What the official platform does not offer is just as important. Messaging and invitation endpoints do exist on paper, but they are locked to approved partners under signed agreements, limited to first-degree connections, and LinkedIn's documentation explicitly requires each message to be tied to a specific member action, ruling out automated or scheduled sending. There is no people-search endpoint for prospecting at all, and no general read access to profiles beyond a member authorizing their own data. If your roadmap says outreach at scale, the official API is not the road.
What are LinkedIn's six API business lines?
LinkedIn organizes its developer platform by business line, not by endpoint, and knowing which line owns your use case is the first step of any application. Consumer covers Sign In with LinkedIn and content sharing, the only truly open corner of the platform. Marketing serves the ads ecosystem: campaign management, audiences, reporting, and conversions for approved marketing partners. Talent powers job postings, Apply Connect, and candidate integrations for HR platforms. Sales is the Sales Navigator Application Platform, SNAP, which embeds Sales Navigator data in CRMs; note that SNAP is not currently accepting new partners, so this door is effectively closed to newcomers. Learning integrates LinkedIn Learning content into LMS platforms. And Compliance exists for regulated industries that must archive member communications.
The pattern across all six: they serve companies integrating with LinkedIn's commercial products, not developers who want programmatic reach into the network. If your use case is prospecting or outreach, no business line covers it, which is the polite structural way LinkedIn says no.
What can you actually do with official access?
Approved partners post and measure ads, sync job postings and candidates, publish company page content, and pull limited member profile data through member consent flows like Sign In with LinkedIn. Worth knowing before you plan around it: LinkedIn currently states it is not accepting new partners for Sales Navigator API access, and the Company Intelligence API is likewise closed to new applications. These are real integrations with real value, and completely separate from the outreach and prospecting features buyers usually mean when they say LinkedIn API.
What messaging and invitation endpoints exist officially?
This is where most developers get their hopes up and then read the fine print. A Messages API and an Invitations API do exist in LinkedIn's documentation. Three restrictions make them unusable for outreach. First, access is limited to approved partners under an API agreement, and the partner list skews to large incumbent platforms. Second, messaging is scoped to first-degree connections: people the member is already connected to. Third, and decisively, LinkedIn requires every message to be tied to a specific member action, and its documentation states that member actions do not include automated or scheduled events. In other words, even an approved partner cannot legally send a sequence.
There is also no people-search endpoint for prospecting anywhere in the official platform. You cannot query the graph for prospects matching a filter, which rules out list building through official channels entirely.
What about the unofficial LinkedIn APIs?
Beneath the official platform sits Voyager, the internal API LinkedIn's own website and mobile apps use, named in LinkedIn's engineering blog but never documented for third parties. Community libraries have wrapped it for years, and the current state of that ecosystem tells you the risk profile: the best-known Python wrapper has had its GitHub repository withdrawn, its last release dates to late 2024, and developers who track the internal endpoints report that LinkedIn rotates its GraphQL query identifiers on deploys, silently breaking integrations. LinkedIn's User Agreement, in its section 8.2 list of don'ts, prohibits scripts, bots, and automated methods to access the services, which is the contractual frame around the entire category.
One nuance worth getting right: LinkedIn does publish an official Python client, linkedin-api-python-client, but it is a thin beta wrapper over the Rest.li protocol for the official APIs above, not a route into messaging or search. The practical unofficial route the market actually uses is different: infrastructure providers that operate member accounts their customers connect, with proxies, pacing, and checkpoint handling built in, which is the category the next section covers.
How do products ship LinkedIn outreach without official access?
Through infrastructure that operates LinkedIn accounts their customers own and connect. The account connects once, and the infrastructure performs the actions a person could perform: viewing profiles, sending connection requests, messaging, InMail. The engineering problems are pacing, warmup, proxies, and staying under LinkedIn's behavioral detection, which is exactly where providers differentiate.
The spectrum runs from raw action APIs, where you build the sequencing and safety logic yourself, to full outreach infrastructure, where campaigns, rotation, safeguards, and the inbox already exist and your product drives them through an API, webhooks, and MCP. Our comparison of Swarmhit and Unipile walks through that trade-off in detail.
Build on outreach infrastructure, not raw endpoints
Swarmhit exposes campaigns, senders, leads, and its unified inbox through an API, webhooks, and an MCP server, with 100+ safeguards monitored 24/7 on every account.
Developer pricing, scoped on a short call.
What does account-based infrastructure actually handle?
Understanding what sits between a raw endpoint and a working outreach feature explains most of the market's pricing. The account layer alone involves a dedicated residential proxy per connected account with stable geolocation, session management that survives LinkedIn's checkpoint challenges (2FA codes, in-app validations, CAPTCHAs), and reconnection flows when sessions drop. Above that sits behavior: warmup curves for new accounts, daily caps per action type, humanized delays between actions, and working-hours schedules per timezone. Above that, the product layer: sequences with branching on accepted or replied, deduplication so two senders never hit the same prospect, reply detection that stops a sequence the moment a conversation starts, and webhooks pushing events into your stack.
Building all three layers is a real product effort; teams that have done it typically report two to three months before campaigns behave reliably, and the maintenance never ends because LinkedIn ships changes weekly. That is the calculation behind buying the engine: the infrastructure absorbs the churn so your roadmap does not.
Which route should AI agents use?
If the consumer of the API is an AI agent rather than an application, the deciding factor changes: agents need the campaign engine exposed as tools, not endpoints to orchestrate. An MCP server over a managed outreach platform gives an agent create-campaign, add-leads, and read-replies as single tool calls, with the safety layer enforced below the tool boundary where the agent cannot accidentally bypass it. Wiring an agent directly to raw account actions inverts that: the model becomes responsible for pacing decisions it has no context to make.
Three routes compared
| Route | What you get | Outreach possible? | Who it fits |
|---|---|---|---|
| Official partner APIs | Ads, jobs, learning, compliance, consented profile data | No: partner-locked, first-degree only, no automated sends | Platforms integrating LinkedIn's commercial products |
| Raw account APIs | Programmatic actions through member accounts you connect | Yes, but you build sequencing, pacing, and safety yourself | Dev teams building an outreach product from scratch |
| Managed outreach platform | Campaign engine, sender rotation, inbox, and safety behind one API | Yes, campaigns as first-class API resources | Teams and products that want outreach working in days |
What does the LinkedIn API cost in practice?
Official access is free: LinkedIn's API Terms of Use state that the APIs are currently provided for free, with the right reserved to charge later. What gates it is approval and a qualifying use case, and some programs carry documented rate caps rather than fees. Third-party infrastructure is priced per connected account: entry points sit around 49 euros per month for small account pools on raw APIs, and full outreach infrastructure is typically scoped to the use case, in the same range at volume. The bigger cost is always the build on top: sequencing, safety judgment, and inbox tooling that raw endpoints leave to you.
Third-party pricing spans a wide range and correlates with how much of the stack is included. Raw channel APIs price per connected account (Unipile, for example, starts at 49 euros per month covering up to 10 accounts). Managed platforms price per sender with the engine included; Swarmhit's API and white-label pricing is scoped on a short call and scales down to $9 per month per sender for large fleets with a 50-account minimum, which lands in the same range as raw APIs at volume, with the sequencer, safety layer, and inbox already built. The honest comparison is not price per account but price per working feature: a cheap endpoint plus three months of engineering is rarely the cheap option.
FAQ
Can I get official LinkedIn API access for outreach?
No. The official programs cover marketing, talent, and approved sales integrations. Messaging and invitation endpoints are partner-locked, restricted to first-degree connections, and explicitly prohibit automated or scheduled sending; people search for prospecting does not exist. Applications framed around outreach automation are not approved.
Is using a third-party LinkedIn API against the rules?
Automation that operates member accounts sits outside LinkedIn's terms, which is true of every tool in the category. The practical risk lives at the account level, which is why pacing, warmup, and behavioral safeguards matter more than any vendor label. Our guide on whether LinkedIn automation is safe covers what actually triggers restrictions.
What is the fastest way to add LinkedIn outreach to my product?
Build on infrastructure where the campaign engine already exists. One API call creates a campaign with senders, sequence, and schedule, instead of months spent wiring raw actions into something reliable.
Is the LinkedIn API free?
Yes. LinkedIn's API Terms of Use state that the APIs are currently provided for free, with the right reserved to charge in the future, and they also prohibit charging your own users incremental fees for access. The cost is the approval gate, not a price list, and some programs carry documented rate caps, such as 500 calls per 24 hours on the Community Management development tier.
Does LinkedIn have a search API for prospecting?
No. The official platform has no people-search endpoint that returns prospects matching filters. Prospecting data comes either from Sales Navigator used manually, from account-based tools operating connected member accounts, or from third-party databases assembled outside the official APIs.
Can AI agents run LinkedIn outreach through an API?
Yes, and it is becoming the default consumption model: an MCP server exposes campaigns, leads, and replies as tools that agents in Claude, Cursor, or custom frameworks call directly, with account safety enforced by the platform underneath rather than left to the model's judgment.
Conclusion
The LinkedIn API question splits on intent. Integrating ads, jobs, or consented member data: apply to the official programs. Building outreach or prospecting features: pick third-party infrastructure deliberately, and weigh raw endpoints against a finished engine. The deeper your feature roadmap, the more the finished engine wins, and our guide on how to build a LinkedIn outreach tool runs the full buy-versus-build math.
For the cost side in detail, including what partner access is worth and how third-party infrastructure is priced, see our LinkedIn API pricing guide.
