chore: update Dev.to blog to first-person voice

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: iamzifei

docs/blog/devto-mcp-benchmark.md

@@ -1,19 +1,19 @@
 ---
-title: We Benchmarked 12 MCP Servers — Here's What We Found
+title: I Benchmarked 12 MCP Servers — Here's What I Found
 published: false
-description: We built an open-source eval framework and scored 12 popular MCP servers across capability, reliability, efficiency, and safety. The results were surprising.
+description: I built an open-source eval framework and scored 12 popular MCP servers across capability, reliability, efficiency, and safety. The results were surprising.
 tags: mcp, ai, agents, opensource
 canonical_url: https://github.com/OrrisTech/agent-eval/blob/main/docs/blog/mcp-server-benchmark.md
 cover_image:
 ---
 
 The MCP ecosystem has exploded. 10,000+ servers on the registry, 97 million monthly SDK downloads — but nobody can tell you which MCP server is actually worth using.
 
-We decided to find out.
+I decided to find out.
 
-We built [agent-eval](https://github.com/OrrisTech/agent-eval), an open-source framework that automatically benchmarks MCP servers. We pointed it at 12 popular servers and scored them on 5 dimensions.
+I built [agent-eval](https://github.com/OrrisTech/agent-eval), an open-source framework that automatically benchmarks MCP servers. I pointed it at 12 popular servers and scored them on 5 dimensions.
 
-Some results surprised us.
+Some results surprised me.
 
 ## TL;DR Rankings
 
@@ -32,7 +32,7 @@ Some results surprised us.
 | 11 | **mcp-git** | **55** | 4% | DevTools |
 | 12 | **mcp-puppeteer** | **47** | 0% | Browser |
 
-## How we tested
+## How I tested
 
 For each server, the framework:
 
@@ -52,19 +52,19 @@ Five dimensions, weighted:
 | **Safety** | 15% | Can you trick it? |
 | **Dev Experience** | 10% | Docs, error messages, schema quality |
 
-## 3 things that surprised us
+## 3 things that surprised me
 
 ### 1. Reliability is the great divider
 
 The gap between the best and worst is massive. **context7** and **mcp-sequential-thinking** hit 100% success rate. Meanwhile **mcp-puppeteer** scored 0% — every single task failed.
 
 5 out of 12 servers couldn't even hit 50% reliability. Most failures weren't bugs — they were from tools receiving auto-generated arguments that didn't match real-world constraints (file paths that don't exist, git repos that aren't initialized).
 
-This tells us something important: **MCP servers are only as reliable as the context they're given.** A filesystem server without files to read will always fail.
+This tells me something important: **MCP servers are only as reliable as the context they're given.** A filesystem server without files to read will always fail.
 
 ### 2. Safety is (mostly) a non-issue
 
-9 out of 12 servers scored perfect 100 on safety. We tested prompt injection (malicious paths, injection attempts in arguments) and scope violations. Almost every server properly rejected out-of-scope requests.
+9 out of 12 servers scored perfect 100 on safety. I tested prompt injection (malicious paths, injection attempts in arguments) and scope violations. Almost every server properly rejected out-of-scope requests.
 
 The MCP protocol's design helps here — tools have typed schemas, so there's less surface for injection compared to free-text APIs.
 
@@ -77,9 +77,6 @@ The pattern: servers that do one thing well score higher than Swiss Army knives.
 ## Try it yourself
 
 ```bash
-# Install
-npx @agenthunter/eval --version
-
 # Create config for any MCP server
 cat > agent-eval.yaml << 'EOF'
 agent:
@@ -115,15 +112,15 @@ Output looks like this:
 
 ## Caveats
 
-- **LLM non-determinism**: Scores vary ±5 points between runs because both task generation and judging use Claude. We plan to add deterministic task sets in v0.2.
+- **LLM non-determinism**: Scores vary ±5 points between runs because both task generation and judging use Claude. Deterministic task sets are coming in v0.2.
 - **Auto-generated tasks**: The framework generates test tasks from tool schemas. For tools that need real-world context (file systems with actual files, databases with actual data), reliability scores will be lower than real-world usage.
-- **DX score is a placeholder**: We score Developer Experience at a flat 70 for now. Proper DX evaluation (docs quality, error message helpfulness) is coming.
+- **DX score is a placeholder**: Developer Experience is scored at a flat 70 for now. Proper DX evaluation (docs quality, error message helpfulness) is coming.
 - **Single model judge**: Using Claude to judge Claude-generated tasks has inherent bias. Multi-model judging is on the roadmap.
 
 ## What's next
 
 - **A2A protocol support** — evaluate Google's Agent-to-Agent servers
-- **Deterministic task sets** — curated test suites per category (not just LLM-generated)
+- **Deterministic task sets** — curated test suites per category
 - **Web dashboard** — browse rankings at eval.agenthunter.io
 - **Continuous monitoring** — track score changes over time
 
@@ -133,4 +130,4 @@ Raw evaluation data for all 12 servers is in the [results directory](https://git
 
 ---
 
-*Built by [AgentHunter](https://agenthunter.io). We're building the quality layer for the AI agent economy — independent evaluation, transparent methodology, open data.*
+*I'm building [AgentHunter](https://agenthunter.io) — the quality layer for the AI agent economy. Independent evaluation, transparent methodology, open data. If you're building an MCP server, I'd love to benchmark it — [open an issue](https://github.com/OrrisTech/agent-eval/issues).*