How to Run a LinkedIn Outreach Campaign Targeting MCP Developers in 2026
Run a targeted LinkedIn campaign for Model Context Protocol (MCP) developers using a proven 3-touch sequence. Copy the messages, refine your list, and send directly from Origami’s built-in sequencer.
Founder @ Origami
Quick Answer
If you’ve already built a list of Model Context Protocol (MCP) developers using Origami — which includes a built-in LinkedIn sequencer — this guide shows you exactly how to refine that list, write a 3-touch campaign with messages you can copy-paste, and launch it all from one platform. No exporting CSVs, no juggling tools.
MCP developers are a specialized, fast-moving audience. They’re deep into agent infrastructure, server tooling, and the evolving protocol itself. Generic outreach will fail. You need messaging that speaks to their world: context windows, tool discovery, server-to-host trust, authentication flows, and the daily friction of building with Anthropic’s protocol. This post gives you the sequence to do exactly that, and demonstrates how Origami lets you go from list → enrich → sequence → send → track without ever switching tabs.
If you haven’t built your list yet, start with our guide on how to build a list of Model Context Protocol (MCP) Developers. Then come back here to turn those names into conversations.
Step 1: Build Your MCP Developer List in Origami
You already have a list, so we’ll keep this brief — but it’s the foundation for everything that follows.
When you log into Origami, you describe your ideal MCP developer in plain English. A good prompt for this audience might be:
"Find MCP developers at AI startups with 10-200 employees who have MCP in their GitHub bio, mention MCP in their job title, or have contributed to MCP open-source projects. Include verified work emails and LinkedIn profiles. Focus on North America."
Origami searches the live web, chains data sources, enriches contacts, and qualifies leads — all from that single prompt. The output is a targeted prospect list containing:
- Verified full name
- Current job title and company
- LinkedIn profile URL
- Work email (and often a direct phone number)
- Company details (size, industry, location)
- Enrichment highlights: GitHub activity, open-source contributions, technology mentions, and MCP-specific signals
You can do this on the free plan (1,000 credits, no credit card required). If you already ran this step and have your list, you’re ready for refinement.
Step 2: Refine and Qualify Your List for LinkedIn Outreach
Not every record that comes back is a perfect fit. Before you fire off a sequence, you need to thin the list so your acceptance rate stays high — and so you’re not wasting touches on contacts who will never reply.
Remove obviously bad fits
Scan the list and cut anyone who doesn’t actually work with MCP. Some profiles might mention “context” or “protocol” in a different context entirely (e.g., network protocols or mobile context APIs). Origami is accurate, but a quick human review is always smart. Look for:
- Job titles that include “MCP”, “Model Context Protocol”, or “Agent Infrastructure”
- GitHub contributions to repos with “mcp” in the name or description
- LinkedIn summaries that mention building MCP servers, MCP hosts, or tool integrations
Segment by role and company type
MCP developers aren’t all the same. Break your list into at least two buckets:
- Server builders: These devs create MCP servers that expose tools (APIs, databases, file systems) to AI hosts. They care about tool definition schemas, authentication, and performance. Their pain points are server-to-host trust, cross-server state, and managing hundreds of tool definitions.
- Host/Client developers: These folks build MCP clients into AI agents, Chat applications, or IDEs. They wrangle multiple MCP servers, combine tool lists, and manage context windows. Their pains are tool discovery, context overflow, and server compatibility.
You can also segment by:
- Company size: Startups (10–50 employees) are more likely to be building their own MCP stack from scratch. Larger companies (50–200) might be integrating MCP into existing agent platforms.
- Location/timezone: If follow-up is time-sensitive, group by region so you can match send times.
- Seniority: Senior or Lead MCP developers often have decision-making authority on tooling, while mid-level devs may be the ones actually implementing.
What “qualified” looks like for an MCP developer
A high-quality lead for a LinkedIn outreach campaign about MCP tooling should hit most of these:
- Has worked directly with MCP (not just “interested in AI agents”)
- Has built or contributed to an MCP server, MCP host, or a tool that integrates with the protocol
- Their company is building on top of MCP (not just experimenting)
- They are in a role where external tooling conversations are welcome (engineering lead, infrastructure engineer, DevTools architect, founder)
Refining the list in Origami is straightforward. You can filter, sort, and tag leads right inside the platform. Once you’re happy, the sequencer is just a click away.
Step 3: Create a 3-Touch LinkedIn Sequence (Copy These Messages)
Origami gives you two ways to build your LinkedIn sequence. You can paste your own templates, set the delay between touches, and launch. Or you can ask the AI agent to generate a personalized sequence automatically, drawing on each lead’s profile data (title, company, industry, GitHub activity) to make every message feel custom.
For a niche like MCP developers, I always recommend writing your own angles first — even if you later let the agent scale personalization. You know the protocol’s pain points better than any generic AI. Below is a full 3-touch sequence you can steal. It’s been run on real MCP developer lists and consistently delivers connection rates above 20% and reply rates in the low double digits.
Cadence: Day 1 (connection request with note), Day 3 (follow-up message), Day 7 (soft close). Adjust delays if you like; 1-3-7 is a proven rhythm.
Touch 1: Connection Request with Note (Day 1)
LinkedIn allows a short personal note when you send a connection request. You get about 300 characters, but shorter messages actually perform better. This note is designed to show you know their work without sounding like a sales pitch.
Hi , saw your work on 's MCP integration for agent tooling. I’m researching how developers handle context management and tool discovery across multiple MCP servers. Would love to connect and swap notes.
Why it works: It calls out the specific tech (MCP integration), names a real pain point (context management, tool discovery), and asks for a peer-level connection, not a demo.
Touch 2: Follow-Up Message (Day 3)
This message goes to everyone who accepted but didn’t reply to the connection note. It’s your chance to open a real conversation. Keep it under 100 words, reference what they’re building, and invite a quick chat.
Hey , thanks for connecting. I noticed your team built an MCP server for — that’s exactly the kind of protocol-level work that’s still rare.
I’m curious how you’re approaching authentication and server-to-host trust. At ${myCompany}, we’re solving that with a proxy that sits between any MCP host and servers. Could I run a quick idea past you?
Variables to set: `` should be auto-populated from Origami’s enrichment data (the specific MCP server or tool they contributed to). If not available, use a generic “your MCP server work”. ${myCompany} is obviously your company name.
Why it works: It shows you did homework, mentions a non-trivial MCP problem (authentication/trust), and frames the ask as an idea exchange, not a sales call.
Touch 3: Final Message (Day 7)
The last touch is a soft close. You’re not asking for anything; you’re planting a flag for when the pain becomes acute. No guilt trips, just value.
Hi , last message, promise. I won’t pitch you again, but I wanted to leave one thought.
As MCP adoption grows, devs are hitting walls with context-window management and server tool discrepancies. We’ve built a layer that normalizes tool definitions across servers and keeps context lean. If that ever becomes a headache, I’d love to chat. If not, no worries — keep pushing the protocol forward.
Why it works: It acknowledges the cadence ("last message"), ties the pain to a concrete technical problem (context window, tool definitions), and positions you as a resource. The tone is respectful, not desperate.
Customization notes: You can replace the specific pain points with your own angle — maybe you’re selling an MCP marketplace, a monitoring tool, or a managed server hosting service. Just make sure every message references something genuinely relevant to an MCP developer’s day-to-day. Generic “we help B2B teams generate leads” will get you ignored.
If you have a larger list (100+ leads), consider letting Origami’s AI agent write personalized variations for each contact. The agent accesses the same enriched data you see (job title, company, GitHub repos, tech stack) and generates unique messages that still follow your strategic angle. It’s a time-saver that doesn’t sacrifice relevance.
Step 4: Send the Sequence Directly from Origami
You’ve built the list, refined it, and written (or chosen) your sequence. Now you send. Everything happens inside Origami — no exporting CSVs, no syncing to a separate LinkedIn tool.
Launching the sequence
From your prospect list, you select the contacts you want to enroll, choose your sequence (the one you built in Step 3), and hit “Launch”. Origami’s built-in LinkedIn sequencer will:
- Send connection requests with your personalized note on Day 1
- Wait the configured delay (e.g., 2 days)
- Automatically send Touch 2 to everyone who accepted but didn’t reply
- Wait another delay and send Touch 3 to anyone still in the sequence
You control the delay between each touch. The platform respects LinkedIn’s safe usage limits to keep your account healthy.
Tracking everything in one dashboard
All activity — opens, clicks, replies — is visible in the same dashboard where you built your list. When you look at a contact’s sequence activity, you also see their full enriched profile: job title, company, GitHub contributions, tools they use, and the original signal that made them an MCP developer. That context is gold. When someone replies, you know exactly why you reached out and what pain point you addressed, so your follow-up conversation starts warm, not cold.
Automatic un-enrollment when a lead replies
This is crucial and often overlooked. If a prospect replies — even with a “not interested” — Origami immediately removes them from the sequence. You’ll never send a follow-up “breakup” message to someone who already responded. You can take the conversation manually, or mark them as a meeting booked, and the platform updates accordingly.
One platform from list to reply
The real advantage: you never leave Origami. List building, enrichment, sequencing, sending, and tracking — all in one place. Other approaches force you to export a CSV from a prospecting tool, upload it into a sequencing tool, and then sync responses back to your CRM. You lose the enriched data that tells you why the lead matters. With Origami, the enriched data is attached to the sequence. When an MCP developer replies, you instantly see they built a server for 1,200 stars on GitHub, work at a 30-person AI startup, and have “MCP infrastructure” in their LinkedIn headline. That’s context that drives better conversations and higher conversion.
Pricing note
The built-in LinkedIn sequencer is included on all paid plans. You’re not paying for sends. You only pay for the credits used to enrich your leads (finding emails, phone numbers, and deep profile data). The free plan gives you 1,000 enrichment credits to test the entire workflow — build a list, run a sequence, and see results — without entering a credit card.
What Results to Expect (and When to Iterate)
For a well-targeted MCP developer list with the sequence above, expect these benchmarks in 2026:
- Connection acceptance rate: 20–30%
- Reply rate (to any message in the sequence): 10–15%
- Meeting-booked rate: 3–5%
MCP developers are technically savvy, skeptical of unsolicited pitches, and very busy. A 3–5% meeting rate might sound low, but if you’re targeting a list of 200 highly qualified leads, that’s 6–10 conversations with people who actually build with MCP. Those conversations can turn into pilots, design partners, or revenue.
If your numbers are significantly below that, don’t immediately rewrite all your messages. First, check your list. Are the contacts really MCP builders, or did a broad search pull in general AI/ML engineers? Refine the list using the qualification criteria from Step 2. A better list beats better copy almost every time.
If the list is solid and acceptance is still low, try a different connection note angle. For MCP developers, you could test a note that mentions a specific open-source contribution or a recent MCP server they shipped. Small tweaks can move acceptance rates by 5–10 points.
If replies are low but acceptance is good, the follow-up message isn’t hitting a sharp enough pain point. Real MCP developers are currently struggling with:
- Tool definition versioning across servers
- Authentication handshakes between hosts and servers
- Context-window overflow when connecting many servers
- Discovery and management of dynamic tool sets
Rotate your Touch 2 message to focus on one of those.