LinkedIn Messaging API for Inbox Access

Summary

• The official LinkedIn API provides no messaging or inbox permission, only w_member_social, OpenID scopes, and Page Messaging for Company Pages.
• The Messages API remains restricted to approved partners with no public application process, and the Compliance permissions that would enable inbox access are closed.
• Third-party APIs such as Unipile address the gap through session-based access (from €49/month), but they operate outside LinkedIn's approved API and carry account restriction risk.
• Kondo is the recommended path for teams that want an AI agent working in their own inbox. Its LinkedIn MCP server reads, drafts, and organizes through the LinkedIn session already open in your browser, so there's no second session on someone else's servers and no third party holding your messages and contacts.

Developers and SaaS builders seeking to read and write LinkedIn inbox messages programmatically face a concrete architectural problem: the official LinkedIn API does not provide this capability, and the partner programs that appear to offer a path forward are closed, non-functional for automation, or inaccessible to most applicants.

The objective is straightforward: synchronize LinkedIn messages into a CRM, stream real-time inbox events into a workflow platform such as n8n or Make.com, or build outreach automation that reads and writes on behalf of a connected user.

What the official LinkedIn API actually provides

LinkedIn's API access framework divides permissions into self-service and partner-gated tiers. Every developer, without approval, can access three capabilities: the OpenID Connect profile and email scopes for Sign In with LinkedIn, and w_member_social for posting, commenting, and reacting on behalf of a member.

No messaging or inbox permission exists in this tier.

The gated partner programs cover Learning, Marketing, Sales Navigator (SNAP), Talent, and Compliance. Personal inbox access does not appear as a product within any of these programs in LinkedIn's own Getting Access documentation.

The Marketing API is the most commonly misread option. The LinkedIn Marketing API terms (last revised July 25, 2025) are explicit: access requires LinkedIn approval, and no right to use any API or data exists without it. The only messaging surface in the entire Marketing API is Page Messaging, which applies to Company Pages, not a member's personal inbox.

The Compliance API is the closest approximation of a sanctioned read-side inbox capability. The r_compliance and w_compliance permissions are engineered for archiving and monitoring member communications — precisely the requirement of many enterprise teams. The documentation leaves no ambiguity regarding availability: these permissions are "listed for reference purposes only. Access is closed and may not be requested."

Even compliance-focused, well-resourced enterprises have no official path to read inbox data via the LinkedIn API.

Why the Messages API is not a usable solution

LinkedIn's Messages API exists but carries a documented restriction: "Usage of this API is restricted to approved partners, subject to limitations via API agreement." However, there is no published application process for becoming an approved partner for this endpoint.

For developers who do reach approved partner status through other channels, the functional restrictions render the API unsuitable for most use cases. As documented by LinkupAPI's analysis of the LinkedIn API, messages can only go to first-degree connections, each message must be initiated by a member action, and nothing can be automated or scheduled.

The LinkedIn API is engineered for applications that extend LinkedIn's own interface, not for applications that run outreach or synchronize inbox data independently of it. Investing engineering time in the partner approval process for an endpoint with these constraints produces a non-functional integration for any inbox automation or synchronization use case.

That leaves three practical ways to give software access to a LinkedIn inbox: third-party APIs, browser agents, and Kondo's MCP server.

How third-party APIs fill the gap

The alternative architecture does not use the official LinkedIn API at all. Third-party providers access LinkedIn by operating from a connected user's authenticated session, using cookie credentials from the browser rather than OAuth tokens issued by LinkedIn's developer platform. This enables read and write access to inbox messages in the same manner as a logged-in member operating through the web interface.

Unipile is one example of this category. Its platform exposes a unified REST API with over 100 endpoints, covering full CRUD capabilities on messages across LinkedIn and other channels including WhatsApp, Instagram, Telegram, and email. For inbox synchronization use cases, it provides realtime webhooks that stream account events as they occur, enabling the "watch my LinkedIn inbox and trigger a workflow" pattern that tools like n8n require.

This architecture resolves the functional problem but introduces a compliance one. Operating via session credentials puts these tools outside LinkedIn's approved API. The risk that developers identify most immediately is account restriction: accounts used for automation can be suspended, sometimes within 24 hours of first use. Third-party providers address this through proxy management, quota controls, and activity pacing designed to keep usage within patterns that resemble normal human interaction. These features reduce the risk of detection but do not eliminate it.

Browser agents

Browser agents such as Claude in Chrome, ChatGPT's agent mode, and Perplexity's Comet operate the LinkedIn website in your own browser, using your own session. Nothing is hosted on a third party's servers apart from what the AI provider sees. The trade-offs are practical. Every search, read, and click the agent makes happens on LinkedIn itself, so working through an inbox produces far more LinkedIn activity than you would on your own, and that activity is visible to LinkedIn. It's also slow across many conversations, long threads consume a lot of your AI plan's usage, and the agent can click send unless you stop it.

Kondo: inbox access for your AI agent, with lower risk

Kondo's LinkedIn MCP server takes a different approach from both hosted-session APIs and browser agents. Like a browser agent, it gives an AI agent (Claude, ChatGPT, Gemini, Copilot, Cursor, or any MCP client) access to your own inbox through the LinkedIn session already open in your browser, with no third party running a session for your account. Unlike a browser agent, it doesn't click through LinkedIn pages. Most of the agent's searches and reads run against messages Kondo already has in your browser or on your computer, so the agent can work across many conversations quickly while adding little activity on LinkedIn.

The agent can list and search conversations, surface unread or unanswered chats, read message history, apply labels, set reminders, archive, and draft replies. Drafts wait in Kondo until you review and send them yourself.

The bigger difference is where the session and data live. A third-party API creates its own LinkedIn session for your account and runs it on its servers, with your messages and contacts stored in its cloud. Kondo works through the session you're already logged into, so no third-party session host stores a copy of your inbox. The only outside service that sees message content is the AI provider you chose to connect, and your agent never sees your LinkedIn credentials.

You also get an inbox to work in, with keyboard shortcuts, labels, reminders, split inboxes, and CRM sync, which an API doesn't provide.

If you would rather reduce the ban risk significantly, this is the safer path — reading and drafting from your own machine, rather than handing session cookies to a third-party host.

LinkedIn DMs Without the API?

Third-party API vs. browser agents vs. Kondo: a direct comparison

For SaaS builders, developers, and teams evaluating the paths, the differences are material.

Access and approval:

• Browser agent: No approval required. Install the extension or agent browser and use LinkedIn while logged in as usual.
• Third-party API: Immediate access by connecting a LinkedIn account through the provider's authentication flow.
• Kondo MCP server: No approval required — sign in to Kondo, connect the MCP server, and your agent reads through your own browser session.

Inbox functionality:

• Browser agent: Reads and acts on the LinkedIn website the way a person would, including sending unless you restrict it. Slow across many conversations.
• Third-party API: Full read/write inbox access, automation, scheduling, and real-time event streaming via webhooks.
• Kondo MCP server: Read and draft access. An agent can list, search, label, remind, archive, and draft replies.

Cost:

• Browser agent: Usually included with a paid AI plan, though reading full pages uses up plan limits quickly.
• Third-party API (Unipile): Published pricing starts at €49/month for up to 10 connected accounts, then €5 per connected account per month for accounts 11 through 50. No additional charge per request or per message.
• Kondo MCP server: $36 for MCP access. See Kondo's pricing for plan tiers.

Rate limits:

• Browser agent: Limited by how fast the agent can load and read pages, and by your AI plan's usage limits. Can easily spike beyond your normal activity pattern.
• Third-party API: Managed by the provider to mirror human activity patterns, typically bundled as quota management.
• Kondo MCP server: Most agent reads and searches run against local data, not LinkedIn. Requests that do reach LinkedIn are paced like one person and back off when LinkedIn signals it is busy.

Session and data:

• Browser agent: Works through your own browser session. Page contents go to the AI provider.
• Third-party API: Runs a separate LinkedIn session for your account on the provider's servers, with your messages and contacts stored in its cloud. Operates outside LinkedIn's approved API, with account restriction risk.
• Kondo MCP server: Works through your own browser session, with no second session and no copy of your inbox on a third party's servers. Message content goes only to the AI provider you connect, and your credentials are never shared.

What to factor into your decision

The recommended starting point for teams that want an AI agent working in their own inbox is Kondo's LinkedIn MCP server. It gives the agent read and draft access through the user's own browser session, without a second session or a third party storing their messages. Browser agents work for occasional tasks, but they're slow across a full inbox. Teams that need full programmatic read/write (sending from code, hosting sessions for client accounts, syncing into a backend, or building an outreach product) are the ones for whom third-party APIs like Unipile remain the relevant, and risk-bearing, option.

Need Inbox Access That Works?

The LinkedIn messaging API landscape is unlikely to open through official channels in the near term. The Compliance permissions are closed with no indication of reopening, and LinkedIn's public documentation shows no new messaging product available to general developers. For teams that need this capability now, the practical options are the ones compared above.

Frequently Asked Questions

Does the LinkedIn API allow reading and writing personal messages?

No. The official LinkedIn API does not grant read or write access to personal member inboxes. Only w_member_social for posting, OpenID scopes, and Page Messaging for Company Pages are available. The Compliance permissions that could support inbox access are closed to new applicants.

Why is there no LinkedIn messages API for developers?

By design. LinkedIn restricts its Messages API to approved partners and offers no public application process. Even those with access face first-degree-only limits and no automation, making it unusable for the inbox sync or outreach tools most developers actually need.

How do I get access to the LinkedIn Messages API?

Through public channels, access is not available. There is no published application process, and questions submitted to LinkedIn's support forums remain unanswered. The endpoint is restricted to approved partners under an API agreement that prohibits automation and scheduling.

What is the alternative to the official LinkedIn API for reading messages?

For teams that want an AI agent working in their own inbox, Kondo's LinkedIn MCP server reads, drafts, and organizes through your own browser session, with no second session and no third party storing your messages. For teams that need full programmatic read/write and accept the restriction risk, third-party APIs like Unipile run a session for the connected account on their servers, enabling full inbox access, real-time webhooks, and automation the official API prohibits. Browser agents such as Claude in Chrome can also read your inbox through your own session, though they're slow across many conversations.

What are the risks of using a third-party LinkedIn API?

These APIs operate outside LinkedIn's approved developer platform, and LinkedIn's User Agreement restricts automated access to the site. The main risk is account restriction, sometimes within 24 hours of first use. Providers use proxy management and activity pacing to reduce detection, but the risk is never fully eliminated. For a closer look at how builders weigh this trade-off, see building with safe LinkedIn automation.