AgentKit Marketing: Automate Lead Research & Outreach for Claude Code (2026)

kit specs and pricing were verified directly on agentkit.best on 2026-08-09; pricing and specs can change, so double-check before you buy.

I'm Jasmine, a developer who uses Claude Code every day, and I actually bought the AgentKit Marketing kit to run lead research for a couple of side projects. This isn't me reading the brochure back to you. I'm going straight at the practical question: how far can AgentKit Marketing actually automate finding and reaching customers, what the workflow really looks like, how the cost compares to SaaS, and when you should not use it.

Which AgentKit? Not OpenAI's AgentKit, and not a lead-gen SaaS

Before we go further, two common mix-ups have to be cleared up - otherwise you'll read this whole piece the wrong way.

Layer 1 - This is not OpenAI's AgentKit. OpenAI also ships a product called "AgentKit" (launched October 6, 2025, made up of Agent Builder + ChatKit + Connector Registry). That one is a platform for building your own agents. This article is about something completely different: AgentKit at agentkit.best (20% off via link) - a bundle of skills and agents that run inside Claude Code / Codex / Copilot through the ak CLI. You don't build agents from scratch; you install the kit and start using it.

Layer 2 - This is not a Clay/Apollo-style lead-gen SaaS. Clay, Apollo, Instantly, and Lemlist are cloud services billed per seat per month, accessed through a web login. AgentKit Marketing is the opposite: it's a kit that runs in your terminal / Claude Code, bought once for $99 (the site lists no recurring fee). Data flows through MCP integrations you connect yourself, and your prompts and output stay on your machine. That's a difference in operating model, not just price - and it decides who this fits and who it doesn't (see the cost and limitations sections below).

Want the big picture first? Read What is AgentKit? A full review. And if you're new to Claude Code, check out what Claude Code is first so nothing feels like it skips a step.

What's inside AgentKit's Marketing kit?

The Marketing kit is one of AgentKit's two main kits (the other is the Engineer kit). Based on what I verified on agentkit.best (2026-08-09), the whole platform ships 108+ skills, 95+ commands, and 45 agents (17 engineer + 28 marketing). The Marketing kit itself focuses on these capability groups:

The sales page describes the capabilities verbatim as "Automate lead research, Scale content creation, Multi-channel outreach" plus "Lead generation automation, Email and programmatic SEO, CRM integrations, Marketing subagents." That's marketing copy - it reads well but it doesn't show you the real workflow, so most of this article is me demonstrating the concrete workflow instead.

An honest note: agentkit.best does not publish the individual name of each agent (like a "Lead Research Agent"), so I describe them by function rather than inventing names. For a closer look at each group, read the detailed Marketing kit review, and to see all 45 agents on the platform, check the overview of AgentKit's 45 agents.

The automated lead research workflow with AgentKit

This is the core. Manual lead research burns the most time on the repetitive part: find, clean, cross-check, score. AgentKit doesn't "invent customers" for you - it speeds up that repetitive chain once you've defined the right inputs. The workflow I actually run has 5 steps:

1. Define your ICP (Ideal Customer Profile). This step is on the human; the agent only assists. I spell it out: industry, company size, decision-maker role, buying signals (for example, "SaaS startups, 10-50 people, recently raised a seed round, need dev tooling"). The more specific the ICP, the less junk in every step that follows.
2. Collect lead sources via MCP. The agent calls the MCP integrations you've connected (public data sources, CRM, or an existing list file) to pull a raw list matching your ICP criteria.
3. Enrich & verify. For each lead, the agent fills in details (website, role, recent signals) and cross-checks against web/CRM MCPs to drop duplicates and stale data. This is where the time savings are most obvious.
4. Classify & score against the ICP. The agent assigns an ICP-fit score with a short reason, so you can prioritize your top leads instead of spreading effort evenly.
5. Compile lead profiles. The output is a table/profile with context (why they fit, hooks for personalization) - ready to hand straight to the outreach step.

The technical key: you don't type 5 separate commands, you invoke one packaged workflow (the "research" group) and let the agent run the chain, pausing at the checkpoints that need your approval. To picture it, a session looks roughly like this (illustrative syntax - the real commands and output depend on your setup):

# Kick off the lead research workflow in Claude Code (illustrative)
ak run marketing:lead-research \
 --icp "SaaS startups, 10-50 people, recently raised seed" \
 --source crm,web \
 --limit 50

# Agent collects -> enriches -> scores, pausing at approval checkpoints
> Collected 74 raw leads -> deduped down to 51
> Enriching 51 leads (website, role, 90-day signals)...
> ICP scoring: 12 A-leads (>=80), 19 B-leads, 20 C-leads
> Exported profiles: ./leads/icp-saas-seed.md (needs your review before outreach)

To dig deeper into how the subagents coordinate along this chain, see the overview of AgentKit's 45 agents.

Personalization & multi-channel outreach (without the spam)

Once you have solid lead profiles, the biggest temptation is to hit "send" on a hollow template to the whole list. That's exactly AI-slop outreach - and it kills both your reply rate and your brand's credibility. AgentKit's real value at this stage isn't "sending faster," it's personalization at scale, grounded in each lead's real data from the enrichment step.

How I use it: the agent pulls the "personalization hooks" from each profile (recent signals, role, company context) and drafts a message tailored to that specific lead, then I review it before it goes out via email/LinkedIn through the matching MCP. A quick comparison:

The difference isn't fancy wording, it's real context (raised seed, hiring devs) pulled from the enrichment step. The agent can scale this because every lead already comes with its hooks attached.

Want to try this exact workflow? The Marketing kit packages the research to personalization to outreach chain so you don't have to wire up each agent yourself. See the AgentKit Marketing kit — 20% off, now $79.20 to match its capabilities against your needs.

Anti-spam compliance (important): automation doesn't mean you get to blast at will. Keep opt-out clear, respect laws like CAN-SPAM/GDPR depending on your market, and don't send high volumes to addresses you have no basis to contact. Done right, automation means more personalization, not more sending.

Setup & a test run (ak CLI + MCP)

Setup is fast - the site says under 5 minutes, and that matched my experience. The ak CLI is a native binary for macOS/Linux/Windows, with no Node or Bun required: the installer auto-detects your OS/arch, downloads the stable build, verifies the SHA-256, then installs to ~/.local/bin. Auth is by email or API key. If you used the old ClaudeKit (ck), migrate with:

# Preview (read-only by default)
ak migrate --from=ck

# Apply for real
ak migrate --from=ck --dry-run=false --yes

After install, you activate the Marketing kit, configure 1-2 MCPs (say, email plus a data source/CRM), and you can run your first research step. I won't repeat the full install process here - that lives in the how to install the ak CLI guide, with each step and the common errors covered.

The real cost: AgentKit vs lead-gen SaaS

This is where AgentKit differs most clearly on the money model. An honest comparison:

Who wins, and when? Solo founders, small SDR teams, and dev-marketers running moderate volume usually come out ahead with the one-time $99 + token model: no fixed per-seat fees, your data on your own machine, and you customize the workflow with Claude Code itself. On the other hand, large teams that need a huge ready-made lead database, multi-seat collaboration features, admin dashboards, and enterprise SLAs are still better served by dedicated SaaS - AgentKit doesn't replace an enterprise CRM. If you're also weighing the Engineer kit, there's a Bundle with both kits at $149.

Want Claude Code stronger right now? If you do moderate-volume lead-gen and want the pipeline running in your terminal instead of another SaaS subscription, the $99 Marketing kit (money-back guarantee) is a cheap entry point to try - just verify the live price before you buy.

See Marketing kit pricing.

Limitations & when NOT to use it

No tool fits everyone. Here are the real limitations I ran into - the sales page won't mention them, but you need to know before you spend:

• You need to know Claude Code first. This is a kit that runs in the terminal/Claude Code, not a click-and-go web app. If you're not comfortable with the CLI and Claude Code, the learning curve is steeper than a SaaS with a UI.
• MCPs need configuring. Lead quality depends on the data sources you connect. Without good MCPs hooked up, even the best agents can only enrich poor data.
• It doesn't replace an enterprise CRM. There's no huge pre-built lead database like Apollo/Clay, and no multi-seat collaboration dashboard. This is an automation tool for your own workflow, not a bundled data warehouse.
• Compliance risk if outreach is careless. Automation makes it easier to over-send. You're responsible for anti-spam and data protection - the tool won't handle that for you.
• Variable token cost. Running many agents over large volumes will burn Claude tokens; factor that into total cost instead of looking only at the $99.

Plainly: if you need a turnkey lead-gen platform for a team of 20 SDRs with a ready-made database, this isn't it yet. If you run a lean process, like control, and already live in the Claude Code ecosystem, it's very much worth a try.

Is AgentKit Marketing worth it for lead-gen? (by persona)

The verdict can't be one-size-fits-all, so here it is by user type:

• Solo founder / indie: Worth it. You trade SaaS seat fees for $99 one-time + tokens, and own the entire workflow. This is the group that benefits most clearly.
• SDR / small sales team (2-5 people): Worth trying, especially if you already use Claude Code. Note the multi-person collaboration side is thinner than SaaS.
• Agency: Depends. Very strong for building repeatable pipelines across multiple clients, but you need someone comfortable with the CLI/MCP at the controls.
• Dev-marketer: A great fit - you have the edge of customizing agents and wiring MCPs exactly the way you want.

If you're still unsure whether AgentKit as a whole is worth it, read What is AgentKit? A full review for a balanced picture of both the engineer and marketing sides before you decide.

Frequently asked questions (FAQ)

How is AgentKit Marketing different from OpenAI's AgentKit?

Completely different. OpenAI's AgentKit (launched October 6, 2025) is a platform for building agents. AgentKit at agentkit.best is a bundle of ready-made skills and agents that run inside Claude Code through the ak CLI - you use them right away, no building agents from scratch.

How does it differ from Clay/Apollo?

Clay/Apollo are cloud SaaS billed per seat per month with a pre-built lead database. AgentKit Marketing runs in your terminal, is bought once for $99 (the site lists no recurring fee), and its data flows through MCPs you connect yourself. Both the operating model and the pricing model are different.

Is automated outreach with AgentKit considered spam?

No, if you do it right. The tool helps you personalize based on each lead's real data instead of blasting a template. You're still responsible for anti-spam compliance (opt-out, CAN-SPAM/GDPR) and not blasting addresses you have no basis to contact.

Do I need Claude Code Pro?

You need Claude Code to run the kit; the right plan depends on your volume. Claude Code has Pro ($20/month) and higher Max plans, or you can use the API billed by token. Running many agents over large volumes will burn tokens, so pick a plan to match your needs.

Is the $99 a one-time or annual charge?

The site shows $99 with no recurring fee listed for the kit, plus a money-back guarantee and lifetime updates for the kit. Since the site doesn't spell out one-time-purchase terms explicitly, verify directly before you buy to be sure.

Do I need to know how to code?

You need to be comfortable with Claude Code and basic command line, but you don't have to be a professional programmer. If you've never used a terminal, start from the ak CLI install guide before running the marketing workflow.

Conclusion & next steps

AgentKit Marketing isn't a "customers on demand" button, but it genuinely automates the heaviest repetitive part of lead-gen - collecting, enriching, scoring, then personalizing outreach - right inside Claude Code, with a one-time $99 model instead of a per-seat subscription. It fits solo founders, small SDR teams, agencies, and dev-marketers already in the Claude Code ecosystem best; it's not the fit yet if you need a ready-made enterprise lead-gen platform. Next step: read the detailed Marketing kit review to scrutinize the capabilities, or start from What is AgentKit? A full review if you want the full picture before deciding.