# CodeMarine

> AI can act. CodeMarine keeps watch.

CodeMarine is an independent protection layer for AI actions. It works alongside provider permissions and frontier models. It does not ask a model to be its own final safety authority.

## Two products

### CodeMarine Guard

CodeMarine Guard is for people using AI desktop applications with real files, accounts and connected tools.

It is being designed to:

- protect files and folders from destructive or overly broad actions
- review sensitive data movement and external destinations
- monitor consequential actions in connected applications
- detect credential collection and exfiltration paths
- keep coverage and limitations visible in plain language

The consumer desktop experience, live proof and broad application coverage are planned. The deterministic action-policy engine already exists. Selected local agent paths are in beta.

### CodeMarine Dev

CodeMarine Dev is for developers and software teams using AI coding agents, editors and pipelines.

It provides:

- local-first deterministic code security
- software supply-chain and dependency analysis
- MCP, skill, rule and agent-configuration scanning
- destructive action evaluation on supported paths
- cross-provider policy and security evidence
- optional frontier-model investigation for selected hard cases

## Product principle

Provider permissions and sandboxes remain important. CodeMarine adds a narrower independent decision around actions that could cause lasting harm.

A preventive claim is valid only when the action passes through a supported pre-execution path. Native, remote and provider-hosted actions remain outside preventive coverage until a separate control path exists.

## Capability status

Current:

- deterministic code and configuration scanning
- software supply-chain scanning
- selected AI artifact scanning
- deterministic destructive-action evaluation
- workspace monitoring

Beta:

- provider adapters for defined shell or write paths
- CodeMarine-mediated MCP operations

Planned:

- CodeMarine Guard desktop experience
- per-surface Protection Center with live proof
- governed MCP and API gateway
- scoped credentials and remote sidecars
- frontier investigation and deterministic patch validation
- organization SSO, SCIM, RBAC and compliance reporting

## Pages

- Home: https://codemarine.ai/
- CodeMarine Guard: https://codemarine.ai/guard/
- CodeMarine Dev: https://codemarine.ai/developers/
- Code Security: https://codemarine.ai/platform/
- Agent Runtime: https://codemarine.ai/runtime-safety/
- Supply Chain: https://codemarine.ai/supply-chain/
- Frontier Models: https://codemarine.ai/frontier-collaboration/
- Developer Integrations: https://codemarine.ai/integrations/
- Trust and Coverage: https://codemarine.ai/trust/
- FAQ: https://codemarine.ai/faq/
- Journal: https://codemarine.ai/blog/
- Documentation: https://codemarine.mintlify.app/get-started/introduction
- Contact: mailto:hello@codemarine.ai