How to Use the WalnutAI MCP Server
Expose WalnutAI capabilities via MCP server. Works with any MCP-compatible client. 10+ tools: gap analysis, test generation, defect patterns, and more. Not locked to WalnutAI surfaces.
MCP: the protocol that matters
MCP (Model Context Protocol) is an open protocol. Claude Desktop uses it. Claude Code uses it. Any AI tool can implement it.
When you build an MCP server, you expose your tools to any client that speaks MCP.
WalnutAI exposes its core capabilities as an MCP server. This means you can use gap analysis, test generation, defect patterns from anywhere.
Not just from walnutai.ai. Not just from Claude.ai. From any MCP-compatible client.
The key insight: no vendor lock-in at the tool level
WalnutAI's proprietary value is the gap analysis algorithm, the test generation model, the defect clustering logic.
We expose that value via an MCP server. You own the integration. You choose the client.
If you want to use Claude Desktop today and switch to a different client tomorrow, your tools still work.
What the MCP server exposes: 10+ tools
WalnutAI MCP server exposes:
gap-analysis: run bidirectional gap analysis on your codebase
test-generation: generate test cases from stories
test-from-defect: generate tests to prevent past defects
defect-analysis: cluster defects, find patterns
code-quality-check: AST-based quality analysis
requirements-trace: build RTM from requirements to code
coverage-report: analyze test coverage gaps
generate-tests-from-docs: test generation from uploaded BRDs
refactor-suggestions: code refactoring opportunities
security-scan: security pattern detection
dependency-analysis: identify dependency risks
More tools added regularly. Each tool follows the MCP protocol. Works everywhere.
Step-by-step: set up MCP server
Step 1: Get your MCP server credentials
Sign into your WalnutAI account at app.walnutai.ai.
Go to Settings > MCP Server.
Copy your MCP server endpoint and API key.
Your credentials: MCP endpoint: https://mcp.walnutai.ai/v1
API key: walnut_mcp_xxxxx...
Keep the API key secret. Treat it like a password.
Step 2: Configure Claude Desktop
Claude Desktop is the easiest entry point. It natively supports MCP servers.
Edit your Claude Desktop config file:
Location: ~/.claude_desktop/config.json {
"mcpServers": {
"walnutai": {
"url": "https://mcp.walnutai.ai/v1",
"apiKey": "walnut_mcp_xxxxx..."
}
}
}
Restart Claude Desktop. The MCP server is now connected.
Step 3: Verify the connection
In Claude Desktop, you should see a new set of tools available in the tool palette.
You can now call any WalnutAI tool directly from Claude.
Example prompt: User: "Run gap analysis on my GitHub repo: acme/payment-service"
Claude: "I'll run gap analysis on your codebase..."
[Claude uses the gap-analysis tool via MCP]
Step 4: Configure Claude Code (optional)
Claude Code also supports MCP servers. Setup is similar:
Claude Code config: Add to your .claudecode/config.json:
{
"mcp_servers": [
{
"name": "walnutai",
"endpoint": "https://mcp.walnutai.ai/v1",
"auth": "bearer walnut_mcp_xxxxx..."
}
]
}
Restart Claude Code. You can now use WalnutAI tools from the IDE.
Step 5: Use any MCP-compatible tool
Your MCP server works with any client that implements the MCP protocol.
Examples:
Claude Desktop: built-in support
Claude Code: built-in support
Cursor: MCP support (configure in Cursor settings)
Continue: open-source code assistant with MCP
Custom tooling: build your own MCP client
Real-world use casesScenario 1: gap analysis in your IDE
You are in Claude Code editing a file. You realize you need to see if requirements are fully implemented.
You run /gap-analysis. Claude Code runs the tool via MCP. Results appear in your editor.
All without leaving the IDE. No context switching.
Scenario 2: test generation in Claude Desktop
You are in Claude Desktop chatting about a feature. You want to generate tests.
You ask Claude to run /test-generation with a specific story.
Claude generates tests, shows them to you, asks if you want to save them.
Scenario 3: integrating with your custom tool
You have an internal tool that needs gap analysis. It speaks MCP.
You configure it to use WalnutAI's MCP server. Now your tool has gap analysis.
No WalnutAI UI needed. No context switching. Just MCP.
Authentication and security Each tool call via MCP includes your API key. The key is used to authenticate the request. Your codebase access (GitHub, GitLab, etc.) is tied to your WalnutAI account. MCP uses the same access permissions. Security checklist: - MCP API key is sensitive. Treat it like a password.
Use environment variables, not hardcoded keys
Rotate your API key regularly
Revoke key access if compromised
Each MCP server call is logged and auditable
Tool availability and status
Not all tools may be available in all MCP clients due to protocol version differences.
Check the WalnutAI MCP documentation for tool availability per client.
Start with the main tools (gap-analysis, test-generation, defect-analysis). Others follow.
Troubleshooting
MCP server is not showing up in Claude Desktop
Check: (1) Config file syntax is correct, (2) API key is correct, (3) MCP endpoint URL is correct, (4) Restart Claude Desktop after config change.
Tool calls fail with authentication error
Check: (1) API key is in the correct format, (2) API key has not expired (rotate if old), (3) Your WalnutAI account is active, (4) API key is not revoked in settings.
Results are slow or timing out
MCP tool calls are synchronous. Large codebases (1M+ LOC) may take 30-60 seconds. This is normal.
If consistently timing out, check your network connection and MCP server status.
Cost: MCP tools use your WalnutAI subscription
Calling tools via MCP counts against your WalnutAI plan usage.
If you have 100 gap analyses/month in your plan, using gap-analysis via MCP counts toward that limit.
No additional MCP-specific costs. Same plan, same limits, new integration path.
The freedom advantage
Traditional integrations lock you into one interface. WalnutAI's MCP server locks you into... nothing.
Your tools work everywhere. You choose your client. You own the workflow.
That is what open protocols enable.
Next steps
Get your MCP credentials. Set up Claude Desktop. Try gap analysis.
Once you see tools working in one client, explore others. Cursor, Continue, custom tooling.
Your WalnutAI capabilities now live in your entire development workflow.
Set up WalnutAI MCP server. Use gap analysis anywhere.https://www.walnutai.ai/
