Sirsi Pantheon

The first product from Sirsi Technologies. Makes AI coding assistants dramatically more efficient — and keeps your infrastructure clean while it works.

brew tap SirsiMaster/tools && brew install sirsi-pantheon

---

Why Pantheon Exists

AI coding tools re-read your entire codebase at the start of every session. On a typical project, 80–90% of that context hasn't changed — wasted tokens, wasted window, wasted money. Your dev machine accumulates invisible waste — ghost app remnants, insecure DNS, orphaned caches. Existing tools miss the hard stuff or require closed-source root access.

Pantheon is different in three ways:

1. AI memory that eliminates cold-start re-reads. sirsi thoth gives AI coding sessions persistent memory via the Model Context Protocol. Thoth compresses your project knowledge into ~2% of the original size — architecture, design decisions, safety rules, recent changes. Your AI starts every session informed, not blank. Works with Claude, Cursor, Windsurf, and any MCP-compatible tool.

2. Ghost detection that nobody else does. sirsi ghosts finds remnants of apps you uninstalled — Launch Services phantoms, orphaned plists, leftover caches, Spotlight metadata for apps that no longer exist. Typically recovers 10–100 GB of invisible waste per machine. Case study: we found 64 GB of Docker VM images the founder didn't know existed →

3. Network security with a safety model borrowed from Kubernetes. sirsi network --fix applies encrypted DNS and firewall hardening using the same probe→mutate→verify→revert pattern as Kubernetes rolling deploys. Auto-reverts within 6 seconds if anything breaks. Case study: the safety model exists because we bricked internet on airline WiFi →

Where This Is Going

Pantheon runs on one machine today. The same scanning architecture — 58 rules, hardware detection, policy enforcement, event ledger — is designed to scale across fleets. Pantheon Ra extends the free tool to multi-machine orchestration: subnet scanning, container auditing, cross-repo AI agents that don't just collect metrics but autonomously detect and fix issues.

The free product solves real problems for individual developers. The enterprise product applies the same intelligence across infrastructure.

	Traditional monitoring (SolarWinds, Datadog)	Pantheon
Architecture	Agents collect metrics, dashboards visualize	Agents scan, detect, and remediate autonomously
AI integration	Bolt-on copilots	Native — persistent memory, MCP server, AI-dispatched orchestration
Privacy	Telemetry to vendor cloud	Zero telemetry. All data stays on your machine.
Ghost detection	Not applicable	Launch Services phantoms, orphaned plists, Spotlight ghosts
Network safety	Alert on misconfiguration	Probe, fix, verify, auto-revert — safely

---

What It Does

Scan for waste

sirsi scan                 # 58 rules across 7 domains — caches, build artifacts, orphaned files
sirsi scan --all           # Deep scan
sirsi scan --json          # Machine-readable output

Hunt ghost apps

sirsi ghosts               # Find remnants of apps you already uninstalled
sirsi ghosts --sudo        # Include system directories

Ghost detection catches Launch Services phantoms, orphaned plists, leftover caches, and Spotlight ghosts that standard cleanup tools miss.

Deduplicate files

sirsi dedup ~/Downloads ~/Documents

Three-phase scan: size grouping → partial hash (8 KB per file) → full hash. Opens a web UI with smart keep/delete recommendations.

System health diagnostic

sirsi doctor               # RAM pressure, disk space, kernel panics, Jetsam events
sirsi doctor --json

Network security audit

sirsi network              # DNS, WiFi, TLS, CA certs, VPN, firewall — read-only
sirsi network --fix        # Auto-apply encrypted DNS + firewall with safety rollback
sirsi network --rollback   # Restore DNS to pre-fix state

The --fix command uses a three-layer safety model: TCP probe before changing config, watchdog polling after, auto-revert within 6 seconds if resolution fails. Case study →

Hardware profiling

sirsi hardware             # CPU, GPU, RAM, Neural Engine, accelerators
sirsi hardware --json      # Full hardware profile

Detects Apple Silicon (ANE, Metal), NVIDIA (CUDA), AMD (ROCm), and Intel. Routes ML workloads to the fastest available accelerator.

AI project memory

sirsi thoth init           # Create .thoth/ knowledge system in your project
sirsi thoth sync           # Sync from source + git history
sirsi mcp                  # Start MCP server for Claude, Cursor, Windsurf

Thoth gives AI coding sessions persistent memory via the Model Context Protocol. Instead of re-explaining your project every session, the AI reads .thoth/memory.yaml and starts with full context.

Code quality governance

sirsi quality              # Full governance audit (coverage, formatting, static analysis)
sirsi quality --skip-test  # Use cached coverage

Runs automatically on every git push via the pre-push gate. Three depth tiers: fast (10-30s default), standard (60-90s), deep (3-5 min pre-release).

Knowledge ingestion

sirsi seshat ingest --source chrome       # Chrome bookmarks + history
sirsi seshat ingest --all-profiles        # All Chrome profiles
sirsi seshat export notebooklm            # Push to Google NotebookLM

Ingests from Chrome, Gemini, Claude, Apple Notes, and Google Workspace. Exports to NotebookLM, Thoth, and Gemini. All data stays local.

---

Install

Homebrew (macOS / Linux)

brew tap SirsiMaster/tools
brew install sirsi-pantheon

From source

git clone https://github.com/SirsiMaster/sirsi-pantheon.git
cd sirsi-pantheon
go build -o sirsi ./cmd/sirsi/

Binary

Download from GitHub Releases.

---

All Commands

Command	What It Does
sirsi scan	Find infrastructure waste (58 rules, 7 domains)
sirsi ghosts	Detect remnants of uninstalled apps
sirsi dedup [dirs]	Find duplicate files with three-phase hashing
sirsi doctor	One-shot system health diagnostic
sirsi network	Network security audit (DNS, WiFi, TLS, firewall, VPN)
sirsi hardware	CPU, GPU, RAM, Neural Engine detection
sirsi guard	Real-time resource monitoring
sirsi quality	Code governance audit
sirsi thoth init/sync	AI project memory
sirsi mcp	MCP server for AI IDEs
sirsi seshat ingest	Knowledge ingestion from browsers and AI tools
sirsi diagram	Generate architecture diagrams (Mermaid/HTML)
sirsi work	Workstream manager — launch AI sessions across projects
sirsi version	Show version and module info

Every command supports --json, --quiet, and --verbose flags.

---

Compatibility

	Supported
AI Assistants	Claude, Gemini, Codex (via MCP)
IDEs	VS Code, Cursor, Windsurf, Zed
Platforms	macOS, Linux, Windows
Architecture	Apple Silicon, ARM, Intel x86

---

Editions

Edition	Scope	Price
Anubis	Single machine — all commands above	Free forever
Ra	Fleet management — multi-repo orchestration, subnet scanning, compliance	Coming soon

The free edition has no feature gating, no telemetry, no time limits. MIT licensed.

---

Security & Privacy

• Zero telemetry. No analytics, no tracking, no data leaves your machine. Non-negotiable.
• Dry-run by default. Every destructive operation requires explicit --confirm.
• 35 protected paths. System directories, keychains, and SSH keys are hardcoded as undeletable.
• Trash-first cleaning. Removals go to Trash with a full decision log.
• DNS safety model. Network fixes probe before changing config, auto-revert on failure.

---

Development

git clone https://github.com/SirsiMaster/sirsi-pantheon.git
cd sirsi-pantheon
git config core.hooksPath .githooks    # Enable pre-push quality gate
go test ./...                          # 1,663 tests across 27 packages
go build ./cmd/sirsi/

See CONTRIBUTING.md for guidelines.

---

Built by Sirsi Technologies

sirsi.ai · GitHub · Pantheon Hub

Apache License 2.0 — free and open source forever.