# OpenMode > OpenMode is a unified AI workspace engineered for cybersecurity, red teaming, reverse engineering, and vulnerability research, featuring uncensored frontier AI models, an autonomous Human-in-the-Loop security agent, token-level usage telemetry, and pay-as-you-go credits. ## Overview OpenMode (https://www.openmode.online/) provides security professionals, vulnerability researchers, and developers with direct, unrestricted access to frontier large language models. The platform features an autonomous security agent capable of executing real-world reconnaissance and analysis tools, backed by granular human approval controls. ## Key Capabilities - **Uncensored Frontier Models**: High-context reasoning, coding, and offensive security models abliterated for cyber research, exploit scripting, and payload synthesis without arbitrary refusals. - **Autonomous Security Agent**: ReAct (Reasoning + Action) execution loop equipped with 8 native security tools and 3 configurable Human-in-the-Loop (HITL) authorization policies (Balanced, Strict, Autonomous). - **Usage Telemetry & Transparent Billing**: Exact token-level deduction logs per prompt and completion. Pay-as-you-go credits powered by PayPal ($0.01 per credit) with 10 free starter credits upon signup. - **Enterprise Security**: Hardware-backed TOTP Two-Factor Authentication (2FA), zero training on user data, strict cryptographic session handling, and user isolation. ## Curated Models & Pricing OpenMode offers 6 curated frontier models with 1,048,576 tokens (1M) context windows: 1. **DeepSeek V4.1 Flash Uncensored** (`deepseek-v4.1-flash-uncensored`): - Tier: MoE High-Speed (FP8) - Pricing: $0.12 / $0.45 per 1M tokens - Ideal for: Sub-second triage, automated exploit scripting, high-volume payload drafting, log ingestion. 2. **DeepSeek V4.1 Flash Thinking** (`deepseek-v4.1-flash-uncensored-thinking`): - Tier: Reasoning Chain-of-Thought (CoT) - Pricing: $0.33 / $1.32 per 1M tokens - Ideal for: Multi-stage logic chains, vulnerability research, complex security bypasses. 3. **Qwen 3.8 27B Uncensored** (`qwen3.8-27b-uncensored`): - Tier: Dense Precision - Pricing: $0.30 / $1.50 per 1M tokens - Ideal for: Deterministic AST code analysis, binary reverse engineering, deobfuscation, shellcode generation. 4. **GLM 5.3 Uncensored** (`glm-5.3-uncensored`): - Tier: Flagship Frontier MoE - Pricing: $0.39 / $1.32 per 1M tokens - Ideal for: Full 1M context decompiled binary analysis (Ghidra/IDA), full-repository audits, OS kernel internals. 5. **GLM 5.3 Flash Uncensored** (`glm-5.3-flash-uncensored`): - Tier: Hybrid Triage (~18B Active) - Pricing: $0.12 / $0.50 per 1M tokens - Ideal for: Interactive triage, PCAP stream parsing, Sysmon event forensics, DFIR investigations. 6. **CyberHeabsy Custom Core** (`cyberheabsy`): - Tier: Offensive Specialist - Pricing: $1.50 / $4.50 per 1M tokens - Ideal for: Dedicated exploit engineering, zero-filter red team operations, proof-of-concept synthesis. ## Autonomous Security Agent Tools The autonomous agent operates using an iterative ReAct reasoning loop with 8 native tools: - `web_search`: Live multi-source web intelligence and query retrieval. - `nvd_cve_lookup`: Direct NIST National Vulnerability Database querying for CVE severity, CVSS scores, and references. - `dns_lookup`: Query A, AAAA, MX, TXT, NS, and CNAME records for target domains. - `dns_recon`: Automated DNS brute-force and subdomain enumeration. - `rdap_whois`: Registration Data Access Protocol lookup for domain ownership and ASN routing data. - `ioc_extractor`: Automated parsing and extraction of IP addresses, hashes (MD5, SHA-1, SHA-256), URLs, and CVEs from raw unformatted text. - `jwt_inspector`: Header/payload decoding, signature verification, expiry checking, and algorithm confusion detection. - `data_decoder`: Multi-layer decoding of Base64, Hex, URL-encoding, binary, and HTML entities. ## Human-in-the-Loop (HITL) Policies - **Balanced (Default)**: Read-only reconnaissance tools (`web_search`, `nvd_cve_lookup`, `ioc_extractor`, `data_decoder`) run autonomously; active queries (`dns_recon`, `rdap_whois`, `jwt_inspector`) require one-click human approval. - **Strict**: Every single tool call requires human review and confirmation before execution. - **Autonomous**: Continuous execution without interruption for fully automated triage and research pipelines. ## Navigation & Canonical Links - Main Workspace: https://www.openmode.online/ - Model Catalog: https://www.openmode.online/models - Autonomous Agent: https://www.openmode.online/agent - Support & Help: https://www.openmode.online/help - Privacy Policy: https://www.openmode.online/privacy - Terms of Service: https://www.openmode.online/terms - Complete LLM Guide: https://www.openmode.online/llms-full.txt