Dunbar Review

Dunbar sells LinkedIn outreach as raw API primitives rather than as software with a dashboard: connect a LinkedIn account, then let an AI agent running in Claude Code, Cursor, Claude Desktop, or ChatGPT send connection requests, message people, and map mutual connections through a credit-metered toolset. This review covers what's actually documented on the live site, what the pricing really costs per action, and what we couldn't verify — including a notable gap between how this tool is described elsewhere and how it describes itself today.

Verdict at a glance

What is Dunbar?

Dunbar (getdunbar.ai, with trydunbar.com redirecting there) is a set of 14 composable LinkedIn actions exposed as an API and MCP server: list and check connections, send connection requests and messages, pull profile and conversation data, search companies and people, and find mutual connections in bulk. There's no campaign builder or lead-list UI — the product is the toolset, and the buyer writes the outreach logic themselves, as prompts or saved workflow files. That "primitives for agents" framing is the whole positioning: built for developers wiring outreach into an AI agent's toolchain, not for a founder who wants to log into a dashboard and launch a campaign.

How it works

Setup: generate an API key, connect a LinkedIn account via OAuth, then point a supported client at Dunbar — a one-line claude mcp add command for Claude Code, a JSON config block for Cursor, OAuth or an API key for Claude Desktop, or a direct OAuth integration for ChatGPT. From there the agent calls the tools directly, with live logging of each action. Outreach playbooks save as plain-text workflow files in Claude Skill format, keeping them portable across MCP clients instead of locked into one campaign builder.

Safety is enforced server-side: a hard 50-sends-per-day cap, 7-day deduplication so the same profile isn't hit twice in a week, a 200-character limit on connection-request notes, and one active batch job at a time. Failed calls refund the credit automatically.

Dunbar pricing (2026)

There's no subscription tier or seat fee — you buy a credit balance and spend it, and credits don't expire. Volume top-ups add bonus credits: $25 gets a 5% bonus, $100 gets 10%, $500 gets 20%.

The real-cost math is straightforward but worth spelling out because the 50-send daily cap is a hard ceiling: running one connected account at the maximum 50 sends/day for a full month tops out at 1,500 credits, or $15 in send costs, before any search or mutual-connection lookups (5 credits/$0.05 each) needed to build the target list in the first place. For a solo operator that's genuinely cheap; it's also a volume ceiling no top-up can raise, since the cap is per profile, not per dollar spent.

What we could and couldn't verify

We found no G2, Trustpilot, or Reddit threads discussing Dunbar under its current name or its trydunbar.com domain — nothing with a meaningful review count, so there's no independent sentiment to report, positive or negative. That's not unusual for a narrow developer tool built on a protocol (MCP) that's still relatively new, but it means prospective buyers are evaluating this on the vendor's own documentation alone.

More notable: material describing this tool elsewhere frames it as a done-for-you outbound service — AI-driven warm intros and multi-channel outreach with human review of every message, reply handling, and meeting booking included, priced from $3,000/month and pitched at B2B SaaS founders as an SDR-hire alternative. The live site describes a categorically different product: a pay-per-credit API/MCP toolkit for developers, no human review step, no meeting-booking service, no pricing above the credit system documented here. We couldn't reconcile the two — it may be a full repositioning of the same company, or the earlier material may be outdated. Confirm directly with the vendor which product you'd actually be buying.

Pros and cons

Pros

• Transparent, granular pricing: 1 credit = $0.01, no subscription or seat fees
• Credits never expire, and failed calls auto-refund
• Native MCP support for Claude Code, Cursor, Claude Desktop, and ChatGPT — plugs into an agent's toolchain without a separate dashboard
• Server-side safety rails (50/day cap, 7-day dedup) reduce the odds of over-sending and getting a connected LinkedIn account flagged
• find_mutuals batch lookups are a genuinely useful primitive for warm-intro targeting at scale

Cons

• No public reviews or ratings found anywhere — effectively unproven as a vendor
• No visible team, company background, or standard pricing page beyond the credit-cost docs
• Requires comfort writing MCP config and your own outreach prompts — no managed campaigns or templates
• The 50-sends-per-day cap is a hard ceiling per account regardless of budget
• Positioning inconsistency with older descriptions of the product that we could not resolve

Dunbar alternatives

Dunbar is LinkedIn-automation infrastructure, not email infrastructure or a sequencer, so the closest comparisons are other LinkedIn outreach tools rather than inbox providers:

• Unipile — unified messaging API (LinkedIn, WhatsApp, email) other SaaS products build on; more established, without the agent-specific framing.
• PhantomBuster — browser-automation LinkedIn scraping and outreach with a longer track record, no native MCP support.
• HeyReach — multi-account LinkedIn outreach built for agencies, dashboard-driven rather than API-first.
• Expandi — cloud-based LinkedIn automation with a visual campaign builder, aimed at non-developers.
• Waalaxy — LinkedIn-plus-email prospecting via a Chrome extension, aimed at solo operators rather than API integrators.

Who should use Dunbar — and who shouldn't

Good fit: developers and technical operators building AI-agent outreach workflows who want metered, per-action LinkedIn primitives and are comfortable writing their own prompts and safety logic; teams already in Claude Code, Cursor, or MCP-compatible tools who'd rather not adopt a separate dashboard.

Poor fit: non-technical founders wanting a managed, human-reviewed outreach service; teams needing more than 50 sends/day from one LinkedIn account; buyers who want a vendor track record or third-party reviews before committing budget.