…llback

Agent log files from Claude Code often have broken parent-child chains
due to missing entries. This caused sidechain matching to fail with
"empty first user or last assistant" errors.

Changes:
- Add AgentID field to LogEntry and ProcessedEntry structs
- Modify getFirstUserMessage to fall back to AgentID-based search
- Modify getLastAssistantMessage with same fallback pattern
- Modify collectSidechainEntries to collect by AgentID when tree
  traversal misses entries

The fix is backward compatible - logs without agentId field continue
to work with existing tree traversal logic.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Document the recommended way to configure the MCP server using `claude mcp add`
for both global and project-local installation, alongside the existing manual
JSON configuration options.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Add a Model Context Protocol (MCP) server that exposes Claude Code session
logs for AI-powered analysis. Key features:

- list_projects: List all Claude Code projects with metadata
- list_sessions: List sessions with time filtering and agent type extraction
- get_session_logs: Get full processed logs for a session
- generate_html: Generate interactive HTML from session logs
  - Supports session_id lookup or direct file_path input
  - Auto-opens in browser when no output path specified
- list_agents: List agent definitions (global and project-specific)
- get_agent_sessions: Find sessions using specific agent types
- search_logs: Search across sessions by content, tool, or role

The generate_html tool mirrors CLI functionality, allowing HTML generation
directly through MCP without needing the CLI binary.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

In newer Claude Code versions, subagent/sidechain logs are stored in
separate files under {session_id}/subagents/agent-*.jsonl rather than
being embedded in the main session JSONL file.

This change updates the parser to automatically detect and load these
subagent files when processing a session:

- Check if {session_id}/subagents/ directory exists
- Load all agent-*.jsonl files from that directory
- Merge subagent entries with main session entries

This fix applies to both CLI and MCP tools since they share the same
parser.ReadJSONLFile() function.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Add 13 new CLI subcommands to match MCP server functionality:
- projects: List all Claude Code projects
- sessions: List sessions for a project
- agents: List agent definitions
- agent-sessions: Find sessions by agent type
- search: Search across sessions
- logs: Get full session logs
- summary: Get session summary
- tools: Get tool usage statistics
- errors: Get session errors
- timeline: Get session timeline
- stats: Get comprehensive stats
- context: Get logs around an entry
- html: Generate HTML from session/file

Key changes:
- Extract Services struct to internal/service/services.go
- Add command infrastructure with Registry pattern
- Support both JSON and human-readable output formats
- Preserve legacy mode (-input flag) for backward compatibility
- Add global flags: --json, --claude-dir, --debug

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Feature/mcp server

Add automated release workflow that:
- Triggers on version tags (v*)
- Builds both cclogviewer and cclogviewer-mcp binaries
- Supports darwin-arm64, darwin-amd64, linux-amd64, linux-arm64, windows-amd64
- Creates GitHub release with all binaries and auto-generated notes

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

All 7 session analysis tools (get_session_logs, get_session_summary,
get_tool_usage_stats, get_session_errors, get_session_timeline,
get_session_stats, get_logs_around_entry) now accept a file_path
parameter as an alternative to session_id, allowing direct analysis
of local JSONL log files without requiring session lookup.

Also adds a Mermaid processing pipeline diagram to the README.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>