AGENTS.md

LOGAN Agent Integration

LOGAN supports AI agent integration through its local HTTP API and MCP server. Any agent can connect to analyze logs and chat with the user.

Quick Start

Your Agent Supports	What to Use	Setup
MCP (Claude Code, Cursor, Windsurf)	MCP tools	.mcp.json already configured — tools auto-discovered
HTTP calls (Copilot, Codex, custom)	REST API polling	Read endpoints below, poll every 2-3s
Persistent connections (scripts)	SSE stream	Connect to /api/events for real-time

Agent-Specific Files

File	Agent	Description
.mcp.json	MCP-compatible agents	MCP server config (auto-discovered)
CLAUDE.md	Claude Code	Project instructions + MCP tool reference
.cursorrules	Cursor	Agent instructions + MCP/HTTP reference
.github/copilot-instructions.md	GitHub Copilot	HTTP polling instructions + endpoints
docs/AGENT_CHAT_GUIDE.md	All agents	Full integration guide with code examples
scripts/agent-chat-poll.py	Any Python agent	Ready-to-run polling chat script
scripts/agent-chat-sse.py	Any Python agent	Ready-to-run SSE chat script

API Reference (http://127.0.0.1:19532)

Chat

Method	Endpoint	Description
POST	/api/agent-register	Register as connected: {"name": "MyAgent"} (shows green dot)
POST	/api/agent-message	Send message: {"message": "text", "name": "MyAgent"}
GET	/api/messages?since=TS	Get messages after timestamp (ms)
GET	/api/events?name=Name	SSE stream (real-time, 1 agent max)
GET	/api/agent-status	Check if an agent is connected

Log Analysis

Method	Endpoint	Description
GET	/api/status	File info, line count, filter state
POST	/api/open-file	Open file: {"filePath": "..."}
GET	/api/lines?start=N&count=M	Read lines
POST	/api/search	Search: {"pattern": "...", "maxResults": 50}
POST	/api/analyze	Full analysis (levels, crashes, components)
POST	/api/navigate	Jump to line: {"line": 123}
POST	/api/filter	Filter: {"pattern": "ERROR"}
POST	/api/clear-filter	Remove filter

Bookmarks & Highlights

Method	Endpoint	Description
GET	/api/bookmarks	List bookmarks
POST	/api/add-bookmark	Add: {"line": 123, "label": "..."}
GET	/api/highlights	List highlights
POST	/api/add-highlight	Add: {"pattern": "...", "color": "#ff0000"}

Baselines

Method	Endpoint	Description
GET	/api/baselines	List saved baselines
POST	/api/baseline-save	Save: {"name": "..."}
POST	/api/baseline-compare	Compare: {"baselineId": "..."}
POST	/api/baseline-delete	Delete: {"baselineId": "..."}

Investigation

Method	Endpoint	Description
POST	/api/investigate-crashes	Crash analysis with context
POST	/api/investigate-component	Component-specific investigation
POST	/api/triage	AI-guided triage

MCP Tools (33 tools)

When connected via MCP, these tools are auto-discovered:

Core: logan_status, logan_open_file, logan_get_lines, logan_search, logan_analyze, logan_filter, logan_clear_filter, logan_navigate

Bookmarks: logan_add_bookmark, logan_remove_bookmark, logan_update_bookmark, logan_clear_bookmarks, logan_bookmarks

Highlights: logan_highlight, logan_remove_highlight, logan_update_highlight, logan_clear_highlights, logan_highlights

Analysis: logan_time_gaps, logan_triage, logan_investigate_crashes, logan_investigate_component, logan_investigate_timerange

Baselines: logan_baseline_save, logan_baseline_list, logan_baseline_compare, logan_baseline_delete, logan_compare_baseline

Notes: logan_get_notes, logan_save_notes

Chat: logan_send_message, logan_wait_for_message, logan_get_messages