Offense Research, Inc.
Offense Research builds and deploys AI agents that safely patch, upgrade, and maintain critical infrastructure for technology companies, starting with GPU clusters. Its Offense product equips coding agents with security tools and structured evidence so they can audit dependencies, inspect code and workflows, and test proposed fixes.
At a Glance
- Technology companies and their infrastructure teams
- GPU-cluster and critical-infrastructure operators
- Software engineering and DevSecOps teams
- Teams using AI coding agents such as Claude Code, Codex, Cursor, Pi, OMP, and Hermes
- +1 more
AI Tools by Offense Research, Inc.
(1)
Offense
Security Toolkit for Coding Agents
Discussions
No discussions yet
Be the first to start a discussion about Offense Research, Inc.
Latest News
Packaged evidence-preserving report handoffs and merged evidence-backed remediation-plan work.
Added bounded port scanning and resumable CLI audits with preview-first scoping.
Released Offense CLI 0.2.0, the first stable versioned open-source CLI release for macOS and Linux.
Added executive PDF reports and Hermes skill support.
Products & Services
Open-source, agent-native security toolkit and optional hosted service. It provides dependency advisory audits, upgrade planning, source and workflow checks, secrets and infrastructure checks, bounded security jobs, structured evidence, and integrations through CLI, MCP, and HTTP.
Stable versioned CLI for macOS and Linux with dependency audits, tool discovery, price quotes, bounded hosted jobs, device login, endpoint-bound credentials, job recovery, optional MCP setup, offline PDF reports, local stack evidence, and Kubernetes inventory previews.
Open-source agent-native policy and approval engine providing deterministic CEL policy and signed, single-use human approvals for AI-agent actions.
Market Position
Offense positions itself as an agent-native security toolkit and platform/harness rather than an autonomous patching bot or a replacement for human reviewers. It combines open-source local analysis, structured evidence, bounded hosted scanners, CLI/MCP/HTTP access, and explicit authorization boundaries; this differentiates it from conventional standalone scanners and turnkey dependency bots while still using established engines such as OSV, Opengrep, Zizmor, Trivy, Gitleaks, Nuclei, and others.
Leadership
Founders
Imran Muthuvappa
Identified in the project's release record as Offense Research's code owner and creator of the remaining logo, icons, illustrations, and social visuals. His published background includes cybersecurity and product/engineering work at Palo Alto Networks, and prior work at Cybrary; he also previously co-founded Kayyo and served as its CTO.
Founding Story
The company says it is working directly with infrastructure teams toward a future in which important infrastructure maintains and secures itself, beginning with GPU clusters. The initial product vision is to give AI coding agents discoverable, bounded security tools and structured evidence while leaving investigation, authorized changes, testing, and approvals with the agent and human operator.
Business Model
Revenue Model
Open-source software with free basic dependency advisory audits and optional hosted security tools sold on a quoted, prepaid, per-tool basis. Customers can also self-host the platform; paid calls require authorization and a spending ceiling.
Pricing Tiers
Basic npm/pnpm dependency advisory audits use public advisory and package metadata and do not require an Offense account or model key.
Catalog pricing is per tool; examples include source-code, dependency, secrets, infrastructure, SBOM, GitHub Actions, OSV, Semgrep, and similar file-based tools.
Catalog examples include $0.50 for HTTPX, $1 for TLS or endpoint discovery, $5 for Nuclei/Schemathesis/Cloudflare checks, $10 for active API security, and $2.50 for live secret verification.
Target Markets
- Technology companies and their infrastructure teams
- GPU-cluster and critical-infrastructure operators
- Software engineering and DevSecOps teams
- Teams using AI coding agents such as Claude Code, Codex, Cursor, Pi, OMP, and Hermes
- Organizations building repository security bots, PR reviewers, secrets guards, CI reviewers, or API testers
- Dependency vulnerability auditing and reviewed upgrade planning
- Security-focused pull-request and source-code review
- Secrets detection and Git-history checks
- GitHub Actions and CI workflow review
- Bounded testing of disposable or verified APIs
- Building custom DevSecOps agents around repositories and organizational approval rules
- Gameplan
- Zown
- Tinkered AI
- Freecode