LogLens — AI Log Investigator for Claude Code

AI Log Investigator for Claude Code

Turn Claude Code into a domain-aware log investigator for your service. Schema-driven flows, pre-filled Loki queries, root-cause analysis — all inside Claude Code's chat. No API key required.

LogLens is a VS Code extension that auto-installs an MCP server for Claude Code. You commit a small .loganalysis/ folder to your repo describing your service's modules and error patterns; teammates install the extension, click one button, and immediately can ask Claude Code:

"Agent says mute is not working, interactionId is abc-123"

Claude calls LogLens, returns ready-to-paste Loki queries, known error patterns, root causes, and a step-by-step investigation playbook.

✨ Features

🚀 Quick Start

Requirements

• VS Code 1.85+ (or Cursor — VS Code fork)
• One of these MCP-aware AI clients installed:
  • Claude Code — reads .mcp.json
  • GitHub Copilot Chat — reads .vscode/mcp.json
  • Cursor (its built-in AI) — reads .cursor/mcp.json
• Node.js 18+ on PATH (used to spawn the MCP server subprocess)

LogLens writes the MCP server entry to all three config locations on Setup, so it works with whichever client you have (or install later).

Installation

1. Install LogLens from the VS Code Marketplace (search "LogLens")
2. Open your service repo in VS Code
3. Click Setup when the activation notification appears
4. Reload VS Code

That's it — 30 seconds. Now talk to Claude Code in chat.

First-time schema generation (schema author, one person)

In Claude Code chat:

"Use LogLens to scan this codebase and generate a starter .loganalysis schema"

Claude scans the codebase, drafts the schemas, asks you to confirm, then writes them. Review the generated files, refine if needed, and commit them to git.

Daily usage (every teammate)

After git pull brings in the committed .loganalysis/:

"Agent says mute is not working — investigate"

Claude returns:

• The correlation ID needed (e.g., interactionId)
• Pre-filled Loki / Grafana query templates
• Specific error patterns to look for with regex + root cause
• Step-by-step investigation playbook

Paste collected logs back to Claude:

"Here are the logs: [paste]"

Claude pattern-matches them against your schema and pinpoints the root cause.

🎯 Why LogLens?

📚 How It Works

VS Code Extension                  Bundled MCP Server                Claude Code
─────────────────                  ──────────────────                ───────────
• Auto-installs MCP server         • Reads .loganalysis/             • Reads .mcp.json
• Detects .loganalysis/ folder     • Exposes 11 tools                • Spawns mcp-server
• One-click setup writes           ▶ list_issue_analyses             • Calls tools in chat
  .mcp.json                        ▶ start_investigation             • Reasons with its
• Schema explorer tree view        ▶ analyze_logs                       own LLM
• NO LLM calls                     ▶ scan_codebase_for_loggers       • Writes schema files
• NO API keys                      ▶ propose_schema_updates             via its edit tools
                                   ▶ ... and 6 more

The extension is a thin auto-installer. All intelligence runs inside Claude Code via MCP tools.

📂 Schema Structure

After setup, your repo has:

.loganalysis/
├── manifest.json                       ← Index of all issue analyses
├── general-context-schema.json         ← Service overview, correlation keys, modules
└── <issue-id>-analysis-schema.json    ← One per issue type (api-errors, auth-failures, etc.)

Commit this folder. Teammates inherit the schemas through git.

Minimal example

// manifest.json
{
  "repo": "my-service",
  "description": "My HTTP API service",
  "generalContext": "general-context-schema.json",
  "issueAnalyses": [
    {
      "id": "api-errors",
      "name": "API Errors",
      "file": "api-errors-analysis-schema.json",
      "keywords": ["error", "500", "timeout"],
      "primaryCorrelationId": "requestId"
    }
  ]
}

Each issue analysis defines: keywords (for matching user descriptions), primary correlation ID, relevant modules, error patterns with regex + root cause, and investigation playbooks.

⚡ Commands

Open the Command Palette (Ctrl+Shift+P / Cmd+Shift+P):

🛠️ Configuration

That's it. No LLM provider settings. No API keys. No models to choose.

💡 Example Prompts

Once set up, here's what you can ask Claude Code:

Investigate an incident:

"Use LogLens to investigate: customer says the call dropped suddenly"

Analyze a log block:

"Use LogLens to analyze these logs and tell me the root cause: [paste]"

Generate or refresh schemas:

"Use LogLens to scan this codebase and propose updated schemas based on recent changes"

Explain a module:

"Use LogLens list_module_files to explain src/payments — what does it do and where can it fail?"

List what's available:

"List all repos and issue categories LogLens knows about"

🔒 Privacy

• No data ever leaves your machine unless Claude Code sends it (your existing Claude Code agreement applies)
• No telemetry — the extension makes zero outbound network calls
• No API keys stored, anywhere
• Schemas stay in your repo — they're plain JSON files you control and commit

🐛 Troubleshooting

Tools not showing in /mcp → Did you reload VS Code after Setup? Ctrl+R in the window. Then check /mcp again.

"No schemas found" → Either your repo doesn't have .loganalysis/ yet (ask Claude to generate one), or the folder is named differently — check loglens.schemaFolder setting.

MCP server crashes → Open a terminal and run:

node "<extension-install-path>/dist/mcp-server.mjs" --workspace "<your-repo>"

Should sit idle waiting for input. If it errors, check the message — most often a schema validation error.

Activation notification doesn't appear → Either you've already dismissed it once, or .mcp.json already has a loglens entry. Run LogLens: Setup MCP Integration for Claude Code manually.

🧰 Requirements

• VS Code 1.85+ or compatible (VSCodium, Cursor)
• Claude Code extension installed and logged in
• Node.js 18+ on PATH (used to spawn the MCP server subprocess)

🗺️ Roadmap

• ✅ MCP server with 9 schema-driven tools
• ✅ VS Code extension auto-installer
• ✅ Multi-repo / monorepo support
• ✅ AI schema generation via Claude Code
• 🔜 Loki / Grafana direct integration (fetch logs from inside the extension)
• 🔜 Datadog / CloudWatch / Elastic backends
• 🔜 Public schema registry for popular frameworks (Spring Boot, NestJS, Django, FastAPI)
• 🔜 Schema versioning + migration tool
• 🔜 Live log tailing during investigation

🤝 Contributing

LogLens is open source under the MIT License. Issues, PRs, and schema contributions are welcome.

• Source: github.com/vikram9410/loglens
• Issues: github.com/vikram9410/loglens/issues
• Discussions: github.com/vikram9410/loglens/discussions

Sharing schemas

If you build a schema for a popular framework, please contribute it to the upcoming public schema registry. One author's effort becomes everyone's onboarding accelerator.

📜 License

MIT © LogLens contributors

🙏 Acknowledgments

Built on top of:

• Model Context Protocol (MCP) — Anthropic's protocol for AI tool integration
• Claude Code — the AI that does the actual reasoning
• Zod — schema validation
• esbuild — bundling the MCP server

Made for engineers who'd rather investigate than google their own error messages.