# August Cyber - Complete Reference (llms-full.txt) > August Cyber (operated by Superneural Technologies Private Limited) provides continuous red teaming, security assessments, security-tuned CyberCART model APIs, scope-locked cloud sandboxes, and the August CLI for engineering teams and autonomous security agents. Website: https://august.so API Base URL: https://api.august.so/v1 Contact: hello@august.so Security Disclosure: security@august.so Mutual NDA: https://august.so/nda --- ## 1. Core Offerings ### Continuous Red Teaming & Security Assessments - Surfaces tested: web applications, REST and GraphQL APIs, authentication and authorization (IDOR, BOLA, privilege escalation, RLS policies), cloud and secrets posture, AI agents, and Model Context Protocol (MCP) servers. - Deliverables: engineer-verified findings with exact HTTP/1.1 request and response evidence, minimal source code patch diffs, verified retesting (200 -> 403 remediation proof), and a signed Letter of Attestation. - Authorization model: no testing begins without a signed Statement of Work and Rules of Engagement specifying exact in-scope staging targets, test accounts, testing window, and emergency stop contacts. Production systems, real customer data, denial-of-service, and phishing are excluded by default. - Sample deliverables: - Sample Red Team Report (PDF): https://august.so/august-cyber-sample-report.pdf - Sample Letter of Attestation (PDF): https://august.so/august-cyber-sample-attestation.pdf - Overview page: https://august.so/sample-report ### CyberCART Model API (`https://api.august.so/v1`) - Security-focused inference gateway supporting three wire formats: - OpenAI Chat Completions: `POST /v1/chat/completions` - Anthropic Messages: `POST /v1/messages` (supports both `Authorization: Bearer` and `x-api-key`) - OpenAI Responses: `POST /v1/responses` - Primary model IDs: `august-cybercart-pro`, `august-cybercart-flash`. - Catalog endpoint: `GET /v1/models`. - Usage and prepaid ledger endpoint: `GET /v1/usage`. - Privacy commitment: zero model training on customer prompts, source code, or findings. ### August Sandboxes (`https://api.august.so/v1/sandboxes`) - Isolated, single-tenant, ephemeral Linux microVMs designed for authorized offensive security testing. - Mandatory ownership verification: before any sandbox can target an external domain, the workspace must prove ownership via DNS TXT record or `https:///.well-known/august-verify.txt` using `POST /v1/targets/challenges` and `POST /v1/targets/verify`. - Scope-locked networking: egress is restricted to verified target domains, private/loopback/metadata IP ranges are blocked against SSRF and DNS rebinding, and command/network evidence is SHA-256 hash-chained (`GET /v1/sandboxes/:id/evidence`). ### August CLI (`august`) & Local MCP Server - Single Rust binary installable via `curl -fsSL https://api.august.so/install | sh`. - Capabilities: - `august auth register` / `august auth verify`: headless email OTP signup and login (no browser required). - `august routes`: local discovery of API routes, auth guards, RLS policies, and MCP tool handlers. - `august sandbox challenge|verify|create|exec|evidence|stop`: full target verification and scope-locked cloud VM lifecycle. - `august attack`: autonomous cross-account authorization probing (local loopback or remote scope-locked sandbox). - `august fix`: minimal source patch synthesis via CyberCART with automatic rollback if retest fails. - `august retest` / `august ci`: byte-exact HTTP/1.1 replay and pull-request CI merge gate (exits 2 if still exposed). - `august mcp`: stdio Model Context Protocol server for Cursor, VS Code, Claude Code, Codex, Zed, and Devin. --- ## 2. Headless Agent Onboarding & Protocols August Cyber implements `auth.md/v1` and RFC 9728 discovery so autonomous agents and coding tools can discover, authenticate, manage scoped keys, and fund workspaces without opening a dashboard: - Protected Resource Metadata (RFC 9728): https://api.august.so/.well-known/oauth-protected-resource - Authorization Server Metadata (`auth.md/v1`): https://api.august.so/.well-known/oauth-authorization-server - Agent Authentication Guide (`auth.md`): https://august.so/auth.md (also at https://api.august.so/auth.md) - Agent Skill Guide (`SKILL.md`): https://august.so/SKILL.md - Agent Skills Index: https://august.so/.well-known/agent-skills/index.json ### Key Scoping & Security Model - Email OTP (`POST /v1/agent/auth/register` -> `POST /v1/agent/auth/verify`) issues standard operational keys (`acrt_...`) with least-privilege scopes (`inference`, `sandboxes`, `targets:verify`, `usage:read`, `account:read`) for new workspaces or verified active workspace owners. - Full-access Admin API keys (`acrt_admin_...`) and privileged scopes (`admin`, `keys:manage`, `billing:manage`, `account:write`) require authenticated web console approval by the workspace owner at https://august.so/console/keys. - Delegated keys enforce strict effective-scope ceilings (no escalation on mint or revoke) and parent-key expiration ceilings. --- ## 3. Canonical Pages - Home: https://august.so - Services: https://august.so/services - How It Works: https://august.so/how-it-works - Sample Report & Attestation: https://august.so/sample-report - Trust & Architecture: https://august.so/trust - FAQ: https://august.so/faq - Mutual NDA (self-serve e-signature): https://august.so/nda - Console: https://august.so/console - Security Disclosure Policy: https://august.so/disclosure - Security Contact (`security.txt`): https://august.so/.well-known/security.txt - Privacy Policy: https://august.so/privacy - Terms of Service: https://august.so/terms