A suite of highly-complementary protocols, design principles, and paradigms for building agentic software.
┌──────────────────────────────────────────────────────────────────┐
│ Agentic Suite │
│ │
│ ┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐ │
│ │ Agentic Repo │→│ Agentic DevOps │→│ Agentic Runtime │ │
│ │ (code) │ │ (deploy) │ │ (operate) │ │
│ │ │ │ │ │ │ │
│ │ SOUL.md │ │ DEVOPS_RUNBOOK │ │ RUNTIME_TOOLS │ │
│ │ skills.md │ │ DEPLOY_SKILLS │ │ RUNTIME_SKILLS │ │
│ │ AGENTS.md │ │ MONITOR_CONFIG │ │ RUNTIME_AGENTS │ │
│ │ TOOLS.md │ │ │ │ │ │
│ └─────────────────┘ └─────────────────┘ └─────────────────┘ │
│ │
│ ┌───────────────────────────────────────────────────────────┐ │
│ │ Shared Foundation │ │
│ │ AMCP (memory/consent) · Policy Engine · Provenance │ │
│ │ Observability · Eval Harness · Framework Adapters │ │
│ └───────────────────────────────────────────────────────────┘ │
└──────────────────────────────────────────────────────────────────┘
| Layer | Module | Purpose |
|---|---|---|
| Agentic Repo | personified_software/openclaw_scaffold/ |
Developer ↔ Code — agents that understand and modify code through conversation |
| Agentic DevOps | agentic_devops/ |
Code → Production — automated build, test, deploy, monitor, and incident response |
| Agentic Runtime | agentic_runtime/ |
User ↔ Running Software — natural-language interface for using any running application |
- Code → Deploy → Runtime: Code changes → DevOps auto-deploys → Runtime refreshes tool registry
- Runtime → Incident → DevOps: Runtime anomaly → auto-create incident → DevOps diagnoses and rollbacks
- Incident → Repo: Post-incident review → auto-create issue/RFC → Agentic Repo assists with fix
Universal OpenClaw-like scaffold toolkit for turning any repository into an agent-ready, personified workspace. Profiles a repo's structure, language, commands, and risks, then generates SOUL.md, skills.md, AGENTS.md, TOOLS.md.
python3 -m personified_software.openclaw_scaffold.cli /path/to/repoDiscovers what a running application can do (API endpoints, CLI commands, etc.) and exposes those capabilities as typed tools for conversational interaction.
python3 -m agentic_runtime.cli profile --app-name my-api --openapi-url http://localhost:8000/openapi.json
python3 -m agentic_runtime.cli generate --profile profile.json --output-dir ./Scans a repository for deployment infrastructure and provides saga-based pipeline orchestration, health monitoring, incident management, and safe rollback.
python3 -m agentic_devops.cli detect /path/to/repo
python3 -m agentic_devops.cli deploy /path/to/repo --env staging --dry-run
python3 -m agentic_devops.cli monitor --target http://localhost:8000/health --type httpMinimal protocol for user-controlled agent memory ownership with consent-based access, cooperative ownership, and portable migration.
python3 amcp/main.py demo
python3 amcp/main.py self-testAll modules share patterns from the design space document:
| Pattern | Used in |
|---|---|
| Typed Tool Invocation | Runtime (ToolManifest), DevOps (PipelineStage) |
| Capability Negotiation | Runtime (ToolRegistry.refresh_from_profile) |
| Planner-Executor Separation | DevOps (PipelineEngine) |
| Saga with Compensating Actions | Runtime (executor.execute_saga), DevOps (pipeline compensators) |
| Human Approval Gate | Runtime (PolicyEngine), DevOps (deployment/remediation gates) |
| Provenance and Attestation | Runtime (ExecutionProvenance), DevOps (PipelineProvenance) |
| Observability Layer | DevOps (MonitorEngine, HealthCheck) |
Both agentic_runtime and agentic_devops provide adapters for:
- PydanticAI — agent with tool functions
- LangGraph — state graph with policy-gated nodes
These follow the same port-adapter architecture as amcp/adapters.py.
# All tests
python3 -m pytest -q
# Per module
python3 -m pytest -q agentic_runtime/
python3 -m pytest -q agentic_devops/