io.github.Kind-ling/twig
MCP server quality scoring and optimization for the agent economy.
Twig
Agents choose you or they don't. Twig makes them choose you.
Part of Kindling — agent SEO for the agent economy.
Free audit. Paid monitoring. Premium intelligence.
| Tier | Features | Price |
|---|---|---|
| Free | twig analyze — score your descriptions. twig optimize — get better variants. | Always free |
| Monitor | Weekly re-scoring, competitive position, intent trend alerts | $0.50–2.00/report via x402 |
| Intelligence | Category benchmarks, intent corpus, competitive index | $0.50–1.00/query via x402 |
| Enterprise | 15% of measured x402 revenue increase | Performance-based, zero if no improvement |
The Problem
Every MCP tool has a description. That description is what an LLM reads when deciding whether to call your service. It's the only signal between "selected" and "skipped".
Most descriptions are garbage.
| Tool | Current Description | Twig Score |
|---|---|---|
exa/answer | "AI-powered answer" | F (28/100) |
wordspace/invoke | "Run wordspace AI agent loop" | F (31/100) |
exa/find-similar | "Find similar pages" | F (33/100) |
jupiter/portfolio | "Wallet portfolio positions" | D (42/100) |
These tools are useful. Their descriptions are not. And getting indexed in x402-discovery or MCPay doesn't help if agents skip you because they can't tell what you do.
Getting indexed is table stakes. Twig helps you get chosen.
Quick Start
npm install -g @kind-ling/twig
# Free — score your descriptions
twig analyze https://myservice.com/.well-known/agent.json
# Free — get optimized variants
twig optimize https://myservice.com/.well-known/agent.json
# Paid — set up weekly monitoring
twig monitor https://myservice.com/.well-known/agent.json
# Paid — see how you rank in your category
twig competitive crypto-defi
# Paid — see what agents are actually searching for
twig intents crypto-defi
Live API
Twig is available as a hosted API at https://twig.kind-ling.com
# Health check
curl https://twig.kind-ling.com/health
# Analyze any MCP server
curl -X POST https://twig.kind-ling.com/analyze \
-H "Content-Type: application/json" \
-d '{"url": "https://your-mcp-server.com/mcp"}'
# Optimize descriptions
curl -X POST https://twig.kind-ling.com/optimize \
-H "Content-Type: application/json" \
-d '{"url": "https://your-mcp-server.com/mcp", "category": "crypto-defi"}'
Response format:
{
"ok": true,
"url": "https://your-mcp-server.com/mcp",
"output": "🔍 Analyzing...\n Score: 72/100 (C)\n ..."
}
No API key required for public endpoints. Rate limits apply.
Scoring Model
Three dimensions, 0–100 each:
| Dimension | Weight | What it measures |
|---|---|---|
| Intent Match | 40% | Similarity to top agent queries for your category |
| Specificity | 35% | Output format, parameters, constraints, examples present? |
| Differentiation | 25% | Do you stand out from similar services in your category? |
Grades: A (85+) · B (70+) · C (55+) · D (40+) · F (<40)
Score below 40 = agents are skipping you.
What Twig Monitors
Your descriptions — re-scored weekly against a live intent corpus. Score changes alert you when your positioning drifts.
Your competitive position — how you rank vs. similar services in your category. Know when a competitor improves and overtakes you.
Intent trends — what agents are searching for this week vs. last. The queries change. Your descriptions need to keep up.
Volume correlation — description changes mapped to x402 volume changes. Close the loop between optimization and revenue.
Optimization Variants
Twig generates 3 variants per tool:
Functional — Action verb + what + return type:
"Answers questions with real-time web sources. Returns a cited response with source URLs, confidence score, and supporting quotes. Best for research and fact-checking."
Structured — Input → output with constraints:
"Web Q&A. Input: question string. Returns: answer (string), sources (URL array), confidence (0-1). Supports: factual queries, current events, research synthesis."
Contextual — Use case → action → output (highest LLM selectability):
"Use when you need a factual answer with sourced citations. Queries the web in real-time and synthesizes a response with URLs and supporting quotes. Best for research, fact-checking, and current events."
Before / After
exa/answer
Before: AI-powered answer ← F (28/100)
After (Contextual): Use when you need a factual answer with sourced citations. Queries the web in real-time and synthesizes a response with source URLs and supporting quotes. Best for research, fact-checking, and current events. ← B (72/100)
wordspace/invoke
Before: Run wordspace AI agent loop ← F (31/100)
After (Structured): Executes a Wordspace multi-step agent workflow. Input: task description and optional context. Returns: completed output, intermediate steps, tool calls made. Best for complex tasks requiring multiple tool invocations. $2/call. ← B (74/100)
Revenue Model
Standard (Monitoring)
Weekly reports are gated via x402 on Base. Set up once, runs automatically.
twig monitor 0xYourWallet --url https://myservice.com/.well-known/agent.json
Enterprise (Performance-based)
For services with existing x402 volume:
twig measure 0xWallet --baseline— record 7-day baseline- Deploy Twig-optimized descriptions
twig measure 0xWallet --compare— measure delta (14–30 days later)- Pay 15% of the revenue increase in USDC on Base
- Revenue flat or down? Pay nothing.
Everything measured via on-chain USDC transfer logs. No self-reporting.
Supported Sources
- A2A Agent Cards (
/.well-known/agent.json) - MCP servers (JSON-RPC
tools/list) - OpenAPI specs
- Local JSON schema files
- x402-discovery catalog (for competitive analysis)
CLI Reference
twig analyze <url> Score descriptions (0-100, free)
twig optimize <url> Generate 3 optimized variants (free)
twig monitor <url> Weekly monitoring (paid via x402)
twig competitive <category> Category ranking (paid via x402)
twig intents <category> Top agent queries (paid via x402)
twig measure <wallet> Measure x402 revenue
twig measure <wallet> --baseline Save baseline
twig measure <wallet> --compare Compare to baseline
twig report <wallet> Revenue report with fee calculation
Categories: crypto-defi, data-feeds, research, computation, media, communication, general
Twig v0.1.0 · Kind-ling · Permanent Upper Class · MIT
Installation
Source-derived launch command. Check the maintainer’s required arguments and credentials before running:
npx -y @kind-ling/twigSet up in your AI client
Merge this template into ~/Library/Application Support/Claude/claude_desktop_config.json. Keep existing servers. Add any arguments, credentials, and permissions required by the maintainer; this template has not been install-tested.
{
"mcpServers": {
"io-github-kind-ling-twig": {
"command": "npx",
"args": [
"-y",
"@kind-ling/twig"
]
}
}
}Restart Claude Desktop completely for changes to take effect. Confirm the server appears connected in the client’s tool list, then try a read-only example from its documentation.
Claude Desktop setup referencePackage
@kind-ling/twignpmCompatible MCP Clients
io.github.Kind-ling/twig works with any MCP-compatible client. Copy the config snippet from the Configuration section above and add it to the file shown for your client, then restart the application.
- Claude Desktop
~/Library/Application Support/Claude/claude_desktop_config.jsonRestart Claude Desktop completely for changes to take effect. - Cursor
~/.cursor/mcp.jsonRestart Cursor for changes to take effect. - VS Code
.vscode/mcp.jsonReload VS Code window for changes to take effect. - Windsurf
~/.codeium/windsurf/mcp_config.jsonRestart Windsurf for changes to take effect. - Claude Code
.mcp.jsonSave at the project root, then start Claude Code in that project and review the MCP server approval prompt. Keep real credentials out of shared files.