How MonkeysCode protects your code, your data, and your infrastructure.

MonkeysCode's cloud services sit behind Google Cloud Armor, Google's WAF and DDoS protection service, running at Google's global edge network — upstream of our infrastructure.

• Web Application Firewall — preconfigured rules covering the OWASP Top 10 (SQL injection, XSS, and other common attacks).
• DDoS protection — always-on defense against Layer 3/4 volumetric and protocol attacks, plus Adaptive Protection: a machine-learning system that detects Layer 7 attacks like HTTP floods.
• Rate limiting — configurable, key-based limits protect the API and dashboard from abusive request volumes.
• Bot & fraud management — integrated with reCAPTCHA Enterprise to detect automated abuse at the edge.
• Geo-based and IP-based access control — allow/deny lists and regional restrictions for compliance.
• Custom security policies — written in Google's Common Expression Language for rules specific to our traffic patterns.

• Encryption in transit and at rest — TLS for all traffic; data encrypted using GCP-managed or customer-managed keys via Cloud KMS.
• Least-privilege IAM — service accounts and roles scoped narrowly per service; no shared broad-access credentials.
• Network segmentation — VPC isolation between the serving plane, the training/RL plane, and the control-plane services.
• Secret management — provider API keys (including BYOK credentials) are envelope-encrypted via Secret Manager and Cloud KMS, never stored in plaintext or in application logs.
• Audit logging — Cloud Audit Logs across infrastructure changes, with Cloud Monitoring and Managed Prometheus for anomaly detection.

• Sandboxed tool execution — every agent action (file edits, shell commands, git operations) runs inside an isolated sandbox — OS-level isolation for interactive sessions, Firecracker/gVisor microVMs for full-autonomous runs.
• Permission governor — side-effecting actions are gated by explicit permission modes (plan / ask / accept-edits / auto), with per-tool allow/deny lists enforceable by org policy.
• Signed, replayable runs — every agent run produces a tamper-evident, signed event log that can be independently verified and replayed.
• Test-gated apply — changes that fail a project's own tests are blocked from landing, by default.

Security here isn't just infrastructure — it's architecture. Where your code goes depends entirely on the model you choose:

• Capuchin (managed) — processed on Colibricode's GCP infrastructure, behind the same Cloud Armor and IAM controls, and never used to train or improve the model unless you explicitly opt in.
• BYOK (Claude, Gemini, ChatGPT) — sent directly to that provider using your own credentials; it never touches Colibricode's servers.
• Local models — never leave your machine.
• Self-hosted / remote models — sent only to the endpoint you configure.
• Local-first indexing — the IDE's codebase index (Merkle tree, embeddings, search) runs on your device by default and is not uploaded unless you explicitly enable cloud sync.

• Self-hosted deployment — run the IDE backend, the model pool, and the extension registry entirely inside your own infrastructure.
• Air-gapped mode — zero outbound network calls, validated end to end, for fully offline environments.
• SSO & provisioning — SAML/OIDC single sign-on and SCIM-based user provisioning.
• Role-based access control — org-level roles, extension allowlists, and version pinning.
• Full audit log — every administrative and agent action logged for compliance review.
• Data residency — choose where data lives: US, EU, LATAM, or fully on-prem.

MonkeysCode's architecture — encryption, access controls, audit logging, and data-residency options — is designed to support common compliance frameworks. SOC 2 Type II audit is planned.

See the Privacy Policy and Terms of Service for our full data protection commitments, including GDPR and CCPA/CPRA disclosures.

We ask security researchers to report vulnerabilities privately before public disclosure, so we can investigate and ship a fix.

We maintain an internal incident response process for security events. In the event of a data breach affecting your information, we will notify affected customers and applicable regulators as required by law — see Privacy Policy §10.

Enterprise and regulated teams can request our security overview, architecture diagram, and audit documentation.