Autonomous · AI Agent Runtime

Your AI. Your Rules. Your Runtime.

A self-hosted AI agent daemon with 19 skill packs, natural language scheduling, multi-tenant workspaces, and threat-aware execution. 7,400+ lines of TypeScript. Runs fully offline.

19 Skill Packs7.4k+ Lines of CodeFully Offline Capable
Core Systems

Not a Chatbot. A Runtime.

Six core systems that transform an LLM from a conversation partner into an autonomous agent that runs, schedules, monitors, and secures itself.

Agent Loop

Autonomous LLM-powered task execution with multi-provider support — OpenAI, Ollama, LM Studio, or any OpenAI-compatible server.

19 Skill Packs

Code, security, devops, data, finance, legal, design, research, and more. Modular capabilities the agent loads on demand.

Natural Language Rituals

Scheduled tasks via cron in plain English. 'Every Monday at 9am, audit the staging server' — Specter handles the rest.

Workspaces

Multi-tenant project isolation. Each workspace gets its own context, history, rituals, and capability grants.

Threat-Aware Mode

Rate limiting, cost caps, anomaly detection, and automatic circuit-breaking. The agent polices itself.

Web Dashboard

Real-time WebSocket updates, session monitoring, ritual status, and workspace management from any browser.

Ecosystem

Ghost Protocol Integration

Specter connects to the full Ghost Protocol stack — persistent memory, secrets management, and automated security assessments.

Wyrm Memory

Persistent memory layer via MCP. Specter remembers context across sessions, learns from past tasks, and builds institutional knowledge.

DragonKeep Security

Secrets management and credential vaulting. API keys, tokens, and sensitive config never touch disk unencrypted.

PhantomDragon Pentesting

Trigger AI-powered security assessments directly from Specter. Schedule recurring pentests as rituals.

Comparison

Cloud Chatbots vs Specter

FeatureChatGPT / ClaudeSpecter
DeploymentCloud-only SaaSSelf-hosted, fully offline
AutonomyResponds to promptsRuns as a daemon, acts on schedule
SchedulingNoneNatural language cron rituals
ChannelsWeb chat onlyTelegram, Web Dashboard, CLI
Data PrivacyData sent to cloudYour data never leaves your infra
Skill SystemPlugin marketplace19 built-in packs + custom skills
Multi-TenancyPer-accountWorkspace-level isolation
Audit TrailBasic logsTamper-evident Event Journal
Architecture

From Config to Autonomous

01

Configure

Define workspaces, grant capabilities, connect your LLM provider — local Ollama or remote API.

02

Deploy

Haunt Daemon starts Specter as a managed process with heartbeat monitoring and PID tracking.

03

Automate

Write rituals in plain English. The agent executes tasks autonomously using its 19 skill packs.

04

Monitor

Web Dashboard shows real-time status. Ghost Replay records every session for debugging and audit.

FAQ

Common Questions

Can Specter run fully offline?

Yes. Connect Ollama, LM Studio, or any OpenAI-compatible local server. No data leaves your infrastructure. Zero cloud dependency.

What are rituals?

Scheduled tasks written in plain English with cron timing. 'Every Friday at 5pm, generate a weekly security report' — Specter parses, schedules, and executes autonomously.

How is it secured?

Capability Engine with token-based access control, Threat-Aware mode with cost caps and rate limiting, tamper-evident Event Journal, and Adversarial Tester for self-fuzzing. v0.7.1 resolved 31 DEFCON-level vulnerabilities.

What's the difference from ChatGPT or Claude?

Specter is a daemon, not a chatbot. It runs on your infrastructure, executes tasks on a schedule, manages its own process lifecycle, and operates across Telegram, web, and CLI — autonomously.

Deploy Your Ghost.

An AI that runs on your terms, on your infrastructure, on your schedule. No cloud lock-in. No data leaks. Just autonomous execution.

Built by Ghost Protocol — autonomous AI for teams who own their infrastructure.