Agent Session Intelligence

Local-first web app to visualize and analyze agent sessions from Claude Code, Cursor, Pi, OpenCode, and Antigravity: context token breakdown, LLM-powered insights, governance pipelines, and skills/rules generation.

الدليل العربي: docs/GUIDE-ar.md

Session data stays on your machine. Nothing is uploaded except LLM API calls you configure for Analysis and Governance.

Agent Session Intelligence — context breakdown

Supported transcript sources:

Claude Code: ~/.claude/projects/**/*.jsonl
Pi: ~/.pi/agent/sessions/**/*.jsonl
Cursor: ~/.cursor/projects/*/agent-transcripts/*/*.jsonl
OpenCode: ~/.local/share/opencode/opencode.db (SQLite) or legacy JSON storage
Antigravity: brain transcripts under ~/.gemini/antigravity-ide/brain/**/.system_generated/logs/*.jsonl, or IDE CLI at ~/.antigravity-ide-cli/projects/*/*.jsonl (additional roots are discovered automatically on Windows/macOS/Linux)

Quick start

Requirements: Bun

git clone https://github.com/ZoheerAlmansoub/agent-session-intelligence.git
cd agent-session-intelligence
bun install && cd web && bun install && cd ..
cp .env.example .env   # optional — for Analysis tab
bun run dev

Platform	Command
Windows	`.\start.ps1`
API + UI	http://localhost:5174 · http://localhost:5173

bun run hooks:install   # optional pre-commit secret scan
bun test                # run unit tests

Features

Context

Treemap, sunburst, and bar charts with drill-down into user messages, tool calls, tool results, attachments, and thinking. Headline stats for input/output, cache read/write, and compaction boundaries.

Messages

Chronological user messages with turn numbers. Copy one or all. Post-compaction filter.

Analysis (LLM)

19 analysis types across six categories — token audit, loop diagnosis, compaction recovery, artifact blueprints, memory drafts, and more. Analysis Pipeline Wizard with optional Apply Pack.

Configure providers in .env or Settings → LLM (live reload, saved locally in .cache/).

Governance

Session and project pipelines (Quick / Standard / Full). Reads AGENTS.md, CLAUDE.md, rules, and skills from disk. Background runs with cancel/resume. Export playbook to docs/governance/.

Artifacts & Insights

Generate skills, rules, hooks, and sub-agent specs. Detect retry loops, tool errors, and token waste across sessions. Insights tab surfaces session patterns; Dashboard aggregates project metrics, eligibility for governance, and cross-session patterns with suggested artifacts.

Each session exposes seven tabs: Context, Messages, Analysis, Artifacts, Insights, Dashboard, and Governance (pipelines, history, playbook export).

LLM providers

Anthropic · OpenAI · OpenRouter · OpenCode Zen · Groq · DeepSeek · Ollama · NVIDIA NIM

See .env.example for configuration.

Security

Never commit API keys. Use .env locally or in-app settings.

bun run check:secrets

Details: SECURITY.md

API

Route	Description
`GET /api/health`	Health check
`GET /api/sessions/:id/transcript`	Full transcript
`POST /api/sessions/:id/analyze`	LLM analysis
`POST /api/sessions/:id/govern`	Session governance
`POST /api/projects/:slug/govern`	Project governance
`GET /api/config/llm`	Provider config (no secrets)

Full reference: docs/AGENT-INTELLIGENCE-ROADMAP.md

Project structure

server/   Bun API, transcript engine, LLM, governance
web/      React + Vite UI
docs/     Roadmap, Arabic guide
scripts/  Tests, secret scanning

Contributing

Contributions welcome — bug reports, transcript parsers, analysis improvements, and docs.

License

GPL-3.0-or-later