Frogo

Production incident investigator CLI that reconstructs cross-service failure timelines and produces a deterministic root-cause hypothesis, with optional MCP-powered log exploration.

Install

npx frogo

Or install globally:

npm install -g frogo

Quickstart

1. Configure integrations and the model:

   npx frogo configure

2. Run the agent chat:

   npx frogo

3. Run deterministic scans:

   npx frogo scan
   npx frogo debug "why did my worker restart?"

Commands

• frogo
  Starts the agent chat (ai-sdk). Uses MCP tools when configured.
• frogo configure
  Interactive configuration for Vercel, Trigger.dev, Datadog, LangSmith, and the LLM provider.
• frogo scan
  Deterministic investigation over the default time window.
• frogo debug "<query>"
  Narrow-window deterministic investigation with a user query bias.
• frogo mcp login langsmith
  Store LangSmith MCP credentials.

Configuration

Frogo reads from:

• Project config: .frogo.json
• Global config: ~/.frogo/config.json

LLM provider

Frogo does not store provider API keys in config. Agent chat uses:

export FROGO_AI_API_KEY="..."

LangSmith MCP

Use the hosted MCP server:

https://langsmith-mcp-server.onrender.com/mcp

Frogo sends your key as the LANGSMITH-API-KEY header.

Set:

export FROGO_LANGSMITH_API_KEY="..."

Datadog MCP

Frogo can connect to a Datadog MCP server via stdio. Provide:

{
  "datadog": {
    "command": "datadog-mcp-server"
  }
}

Set:

export FROGO_DATADOG_API_KEY="..."
export FROGO_DATADOG_APP_KEY="..."

Philosophy

• Deterministic pattern engine first
• Deterministic scans do not require LLM output
• Normalized events only (no raw logs sent to the model)
• Built to extend with more connectors and patterns

License

MIT