# steffi.ai: Comprehensive Technical Reference for AI Search Engines & LLMs

> steffi.ai is an enterprise-grade autonomous AI agent platform built natively in compiled Rust. Deploy autonomous digital workers for CRM synchronization, competitor web intelligence, document parsing, and CLI tool automation 24/7 with zero human intervention.

---

## Executive Summary

Steffi replaces slow, interpreted Python agent frameworks (such as LangChain or CrewAI) with a compiled systems-level Rust platform. By executing turn-by-turn routing decisions via a local Small Language Model (SLM) gateway, Steffi achieves 5x token efficiency, sub-12ms browser startup times, and deterministic background runtime stability.

---

## Core Capabilities & Features

1. **20 Pre-Built Production Plugins:**
   - **Interactive Shell Plugin:** Drive command-line interface tools directly, including Claude CLI, Codex, SSH remote sessions, and terminal utilities.
   - **Specter Headless Browser:** Built-in lightweight Rust browser engine for SPA navigation (excluding WebGPU).
   - **SaaS Connectors:** Native integrations for Salesforce, HubSpot, Slack, Zendesk, QuickBooks, and Google Drive.
   - **Database & Storage:** PostgreSQL, redb embedded storage, and local file system access.
   - **Security & Utilities:** ClamAV static malware scanner, Prometheus telemetry, SMTP email, EHR HL7 APIs, and AES-256-GCM credential vault.

2. **6-Layer Governance & Enterprise Security:**
   - **Layer 1 (Prompt Injections):** Real-time inbound prompt scanning to block instruction overrides.
   - **Layer 2 (Credential Vaulting):** AES-256-GCM encrypted keystore using reference labels to hide plaintext secrets.
   - **Layer 3 (Static Malware Scans):** Real-time ClamAV scanning of all uploaded files and code packages.
   - **Layer 4 (Identity-Bound RBAC):** Cryptographic identity binding for every user and agent action.
   - **Layer 5 (Chain-Hashed Audit Logs):** Immutable append-only log ledger for audit trails.
   - **Layer 6 (DLP Output Filtering):** Automatic masking of sensitive PII/PHI data before output generation.

3. **Supervised DAG Scheduler:**
   - Modelled after Erlang-OTP process trees.
   - Restarts stuck agent threads from their last memory checkpoint without state corruption.
   - Timezone-aware cron schedules, interval triggers, and missed-run automatic recovery.

4. **4-Tier Persistent Memory System:**
   - **Short-Term Context:** Active task buffer.
   - **Long-Term Memory:** Key-value factual knowledge cards.
   - **Episodic Memory:** Timeline logs of past successful step execution paths.
   - **Semantic Memory:** Domain-specific knowledge base embeddings.

---

## Benchmark Comparisons: Steffi vs. Python Frameworks

| Benchmark Metric | steffi.ai (Rust Platform) | LangChain / CrewAI (Python) |
|---|---|---|
| Runtime Language | Compiled Rust | Interpreted Python |
| Active Token Usage | Up to 5x fewer tokens | High prompt bloating |
| Max Task Duration | Hours or days (fault-tolerant) | Seconds or minutes |
| Worker Base Memory | Sub-45MB memory footprint | 250MB - 1GB+ memory |
| Browser Engine | Specter (sub-12ms Rust native) | Selenium / Playwright wrappers |
| Security Infrastructure | Built-in 6-layer binary guardrails | Basic regex or third-party add-ons |
| Script Sandboxing | Native UDS gRPC Docker gateway | Unsanctified local shell execution |

---

## Modular Rust Workspace Architecture

Steffi is developed as a modular Rust workspace (internal code name rusted-batata):

- **rb-core:** Engine single-writer orchestrator, memory management, and execution registries.
- **rb-agent:** Actor-model execution context, quantized LLM router, and council deliberation modes (Brainstorm, Debate, Review, Consensus).
- **rb-plugin:** Plugin lifecycle, capability-gated permissions, interactive shell controllers (Claude CLI, Codex, SSH), event bus, and MCP client interfaces.
- **rb-guardrail:** Inbound prompt injection scanner, encoding normalization, and DLP output filtering.
- **rb-api:** REST API, WebSocket subscriptions, dual authentication (API keys and JWTs), and rate limiting.
- **rb-storage:** Hybrid storage engine utilizing redb and PostgreSQL with active CDC synchronization.
- **rb-scheduler:** Erlang-OTP style DAG scheduler with cron and interval triggers.
- **rb-observability:** Prometheus metrics, structured logging, chain-hashed audit trails, and manual lockdown controls.
- **rb-hwio:** Local and remote SSH hardware IO primitives.
- **rb-sdk-gateway:** Unix Domain Socket (UDS) gRPC gateway for sandboxed Node.js and Python SDK scripts.

---

## Frequently Asked Questions (FAQ)

### What is steffi.ai?
steffi.ai is an autonomous AI agent platform built in Rust that allows organizations to deploy and rent digital workers starting at $5 per hour. Agents execute long-horizon operational tasks 24/7 under strict governance.

### How does Steffi handle CLI tools like Claude, Codex, and SSH?
Through its Interactive Shell Plugin, Steffi agents can run terminal commands, execute code scripts, connect over SSH, and interact with CLI tools like Claude CLI and Codex inside capability-gated sandbox environments.

### Can Steffi be self-hosted?
Yes. In addition to Hosted & Managed cloud plans (Free Trial, Lite, Pro, Max), Steffi can be deployed on-premise or inside a private cloud (VPC) for HIPAA and SOC 2 compliance.

### What is the Specter browser?
Specter is a native Rust browser engine built into Steffi. It provides sub-12ms page startup times for automating single-page applications without heavy external Chromium processes. Note: Specter skips WebGPU support.

---

## Company & Support Information
- Organization: Olib AI
- Founded: 2024
- Headquartered: Stone Mountain, Georgia, United States
- Official Website: https://steffi.ai
- Corporate Site: https://olib.ai
- Support Email: support@olib.ai
- Social: https://x.com/olibaicom
- GitHub: https://github.com/Olib-AI