integration.md

Integration Guide

Saguaro is designed to be the "Cortex" for other AI Agents. This guide explains how to integrate Saguaro into your agentic workflow.

1. CLI Usage (JSON Mode)

All Saguaro commands support structured output, making them easy to parse.

Retrieval

saguaro query "database connection pool" --json

Output:

{
  "matches": [
    {
      "file": "src/db/pool.py",
      "score": 0.89,
      "type": "class",
      "line": 45
    }
  ]
}

Perception (SSAI)

Use the Standard Agent Interface (SSAI) to read code.

1. Get Context: saguaro agent skeleton path/to/file.py
2. Read Function: saguaro agent slice ClassName.function

2. Model Context Protocol (MCP)

Saguaro exposes an MCP-compatible server.

saguaro serve

This starts a stdio or SSE server that exposing resources:

• saguaro://query?q=...
• saguaro://file/...

3. Python API

For deep integration, import the saguaro package directly.

from saguaro.indexing.engine import IndexEngine

# Load existing index
engine = IndexEngine(repo_path=".")

# Vector Search
results = engine.search("authentication", k=5)

for res in results:
    print(f"{res.file}: {res.score}")