DeAgent

Security

Built for teams who can't afford mistakes

Safety controls with full transparency. Inspect the code, self-host on your infrastructure, and bring your own keys.

Never delete. Never destroy. Notify instead.

Destructive operations are blocked. Your team is notified before anything risky happens.

Open source. Verify it yourself

DeAgent is MIT licensed. Your security team can audit every collector, tool, and safety check.

  • ·MIT licensed. Free to use, modify, and self-host
  • ·Full source code available on GitHub for security review
  • ·Bring your own keys. No vendor lock-in for AI or cloud credentials
  • ·Security policy and guarantees documented in the repo

BYOK architecture

Each org stores their own Anthropic, Slack, GitHub, and cloud keys. Platform operators run shared queue/storage only — never tenant secrets.

Org agent → org credentials → their infra

Platform → shared storage/queue only

  • MIT license — inspect everything
  • Every collector and safety rule on GitHub
  • BYOK — org-scoped credentials, never shared
  • SECURITY_GUARANTEES.md — destructive ops blocked
Read SECURITY_GUARANTEES.mdInspect collectors and safety rules on GitHub →

GitHub metrics

Live repository activity: contributors, commits, and project health signals your team can verify before deployment.

2

Stars

4

Forks

12

Open issues

4

Contributors

43

Total commits

Last push: 7/22/2026, 9:13:15 PM

Contributor breakdown

View graph
OkiriGabriel
@OkiriGabriel38 commits (88%)
Nitjsefnie
@Nitjsefnie3 commits (7%)
Maqbool61
@Maqbool611 commits (2%)
kevinnft
@kevinnft1 commits (2%)
open-devops-agent / open-devops-agent

Open-source codebase for DeAgent

mainMIT★ 2
CodeBranchesContributors

Branch strategy

  • main: Default branch. Stable, production-ready code
  • feat/*: Feature branches for new capabilities
  • fix/*: Fixes, hardening, and security patches
View repositorySecurity docs

Never delete production data

All destructive operations are permanently blocked. The agent sends email alerts instead of executing delete or destroy commands.

Multi-layer authentication

API key auth, webhook HMAC signatures, IP whitelisting, and optional VPN/bastion integration.

Complete audit trail

Every decision, action, and escalation logged to org-scoped cloud storage (S3, MinIO, GCS, Azure Blob).

RBAC & approval gates

Role-based access control with Slack approval required for rollback, scale-down, delete, and exec actions.

Security principles

  • Dry-run by default. Test before executing
  • Command blacklist prevents dangerous operations
  • Emergency stop switch disables all agent operations
  • PII scrubbing in logs and notifications
  • Security scanning and compliance checks (CIS, SOC2)
  • No shell injection. Safe command execution via shlex
  • Credentials encrypted at rest, API keys never logged
  • Rate limiting on all API endpoints
Talk to our teamRead the docs