Open Source
Open where trust matters.
The components that determine whether Guard is trustworthy are open. The enforcement engine, the audit trail, and the CLI are Apache-2.0 — readable, auditable, and forkable.
4 open-source components
These are the exact components open-sourced as of the last audit (2026-09-01). All four are Apache-2.0. The hosted product (SaaS at conductai.ai) runs the same code.
| Component | Licence | Repository path | Purpose |
|---|---|---|---|
| conduct-cli PyPI: conduct-cli | Apache-2.0 | packages/conduct-cli/ | Agent lifecycle management, Guard sync, policy testing. Runs on developer machines and in CI/CD. |
| Guard runtime core | Apache-2.0 | apps/api/app/guard/ | Core enforcement engine: rule evaluation, decision scoring, hash-chained audit trail. Runs inside the Conduct API server. |
| Playbook DSL compiler | Apache-2.0 | apps/api/app/compiler/ | Compiles YAML playbook definitions into executable DAG runs. Shared by every shipped playbook. |
| Agent Booster MCP | Apache-2.0 | tools/booster/ | Developer productivity tools (semantic search, smart read, test coverage) for Claude Code and Cursor. Runs locally or on Vercel. |
conduct-cli
Apache-2.0packages/conduct-cli/
Agent lifecycle management, Guard sync, policy testing. Runs on developer machines and in CI/CD.
PyPI: conduct-cli
Guard runtime core
Apache-2.0apps/api/app/guard/
Core enforcement engine: rule evaluation, decision scoring, hash-chained audit trail. Runs inside the Conduct API server.
Playbook DSL compiler
Apache-2.0apps/api/app/compiler/
Compiles YAML playbook definitions into executable DAG runs. Shared by every shipped playbook.
Agent Booster MCP
Apache-2.0tools/booster/
Developer productivity tools (semantic search, smart read, test coverage) for Claude Code and Cursor. Runs locally or on Vercel.
What open means here
Apache-2.0
The entire repository — CLI, Guard runtime, compiler, web UI, and API — is Apache-2.0. The SaaS product runs this source. There is no closed core.
Self-hosted
Run Guard on your own infrastructure using Docker Compose. The same policy engine and audit trail, on hardware you control. Kubernetes support is in preview.