While designing a new subsystem for Voyant.io, I was working through a scaling problem with Claude. We had about 12k signals in PostgreSQL, but I knew that wouldn't hold up as we grew. Claude surprised me with this:
High-Volume Signal Storage
PostgreSQL works fine for 12k signals but won't scale to millions.
Solution: You have ScyllaDB MCP - perfect for time-series signals
- Partition by org_id + date
- TTL for automatic cleanup
- Sub-millisecond reads at scale
"You have ScyllaDB MCP." Claude didn't just recommend a database—it noticed I had the MCP server installed and made a specific, contextual recommendation based on my exact situation.
This is an emerging channel for developer tool distribution—and it's worth paying attention to.
SEO Still Works (But Maybe Not Forever)
I've written before about how GEO depends on SEO. Traditional content strategy continues to work in the agentic era because AI agents often access content through search indexes. Your blog posts, documentation, and landing pages still matter.
But there's a hierarchy emerging for how agents find and recommend tools. And MCP servers sit at the top of it.
The Context Hierarchy
We've written before about how search indexes are becoming DNS for agents—the discovery layer that routes AI assistants to capabilities. But there's a hierarchy here worth understanding:
Tier 1: MCP Servers on the Developer's Machine
Highest priority. Your tools are literally in the agent's context window. The agent
understands your capabilities deeply and can speak to them credibly.
Tier 2: Training Data
What the model learned during training. Useful, but static, months old, and shared
with competitors who appear in the same documentation.
Tier 3: Web Search Results
Agents may search the web for context, but this is filtered through RAG systems
and often summarized.
When Claude recommended ScyllaDB to me, it wasn't just pulling from training data. The MCP server was in context, so Claude could see exactly what ScyllaDB offers and explain why it fit my use case.
The Contextual Advantage
Here's the crazy part: even "Generative Engine Optimization" is built on SEO. When agents need context, they often pull from search indexes. So all that GEO advice? It's still playing the same game—just with extra steps.
MCP is different. With an MCP server on the developer's machine, your technology—with all its opinionated, highly differentiated capabilities—is sitting right there in the agent's context. No search ranking. No summarization. The agent can speak credibly to what makes you different because it can see your tools, understand your API, and make specific recommendations.
That's not marketing copy. That's context.
Why This Matters
The ScyllaDB MCP server case study explored how MCP changes the discovery dynamic. When an agent has your tools in context, it can give specific, credible recommendations rather than generic summaries.
Instead of "ScyllaDB is a high-performance NoSQL database," the agent can explain why it fits the developer's specific requirements—because it understands the capabilities from the MCP tool definitions.
This is a different kind of recommendation. Not a search result. Not a summary of your marketing page. A contextually relevant suggestion from an agent that actually understands what your product does.
A Complement to PLG
Product-Led Growth remains the dominant GTM strategy for developer tools. Make it easy to try, reduce friction to adoption, let the product sell itself.
MCP servers are a complement to this, not a replacement. They add another channel—one where agents can recommend your product credibly because they understand it deeply.
Think of it as giving agents the context they need to be good advocates for your product. They're not replacing your docs or your free tier. They're adding a layer where your differentiation can shine through in the moment a developer asks for help.
What You Should Do About This
1. Build an MCP Server for Your Core Capabilities
Not a wrapper around your API. A thoughtfully designed set of tools that let agents demonstrate your differentiation. What queries should an agent run to prove your performance claims? What comparisons would make your competitive advantages obvious?
2. Make Installation Trivial
The developer has to install your MCP server for any of this to work. One command. No configuration hell. If it takes more than 30 seconds, you've lost.
3. Include Competitive Comparison Tools
The ScyllaDB MCP server doesn't just let agents query ScyllaDB—it lets them compare ScyllaDB to DynamoDB head-to-head. Don't be afraid of comparison. If your product is genuinely better, let the agent prove it.
4. Instrument Everything
You need to understand how agents are using your MCP server. Which tools get invoked? What patterns lead to adoption? This is the new analytics—agent behavior, not page views.
The Window Is Open
Most developer tool companies haven't added this channel yet. They're doing the right things—good docs, solid content strategy, SEO—but they're missing the emerging layer where agents make contextual recommendations.
The companies that build compelling MCP servers now will have a head start as agent-mediated discovery grows. It's not about abandoning what works—it's about adding a channel where your differentiation can really shine.
Ready to Make Your Product AI-Discoverable?
We help developer tool companies build MCP servers and implement AI-native distribution strategies.
Schedule a Demo