Love OpenClaw? This is what comes next. A production-grade AI agent infrastructure that compiles into a single Rust binary with auth, memory, MCP hosting, and observability built in. One dependency: PostgreSQL.
Built on systemprompt.io. You own the code. You extend it. You deploy it anywhere.
This is a library, not a framework. No vendor lock-in. Self-hosted with PostgreSQL. Migrating from OpenClaw?
Quick Start
Prerequisites
- Rust 1.75+: rustup.rs
- just: command runner (
cargo install just) - Docker: for local PostgreSQL (docker.com)
1. Create Your Project
Click "Use this template" on GitHub, or run:
gh repo create my-agent-platform --template systempromptio/systemprompt-template --clone
cd my-agent-platform
2. Build
just build
3. Login & Setup
just login # Authenticate with systemprompt.io cloud
just tenant # Create your tenant
4. Start
just start
Visit http://localhost:8080
What You'll See
When you open the homepage, you'll see the "Welcome to the endgame" dashboard:
┌─────────────────────────────────────────────────────────────────┐
│ Welcome to the endgame │
│ │
│ Agents that regenerate, learn and orchestrate agents. │
│ Run by superagents like Claude Code. Directed by you. │
│ │
│ ┌──────────────────────────────────────────────────────────┐ │
│ │ systemprompt online │ │
│ │ ├── MCP Services: 3 running │ │
│ │ └── Active Agents: 1 ready │ │
│ └──────────────────────────────────────────────────────────┘ │
│ │
│ [Level 1: Chat] [Level 2: Tools] [Level 3: Agents] [Level 4] │
│ │
└─────────────────────────────────────────────────────────────────┘
Each level card shows runnable commands you can copy and execute to experience that level of AI capability.
The AI Evolution
The homepage demonstrates the four levels of AI capability. This template takes you from Level 1 to Level 4:
Level 1: Chat / The Copy-Paste Era
Single-turn conversations. You ask, AI answers, you copy and paste.
systemprompt admin agents message welcome -m "Hello!" --blocking
Level 2: Tools / The Function-Calling Era
AI invokes tools. It reads data, calls APIs, writes files.
systemprompt plugins mcp tools
systemprompt plugins mcp call content-manager create_blog_post -a '{"skill_id": "announcement_writing", "slug": "welcome-post", "description": "AI tools demo", "keywords": ["mcp"], "instructions": "Write a welcome announcement."}'
Level 3: Agents / The Autonomous Era
Define the goal, agents find the path. Multi-step reasoning with tool use.
systemprompt admin agents message welcome -m "Create an announcement blog" --blocking --timeout 120
systemprompt infra jobs run publish_pipeline
Level 4: Mesh / The Orchestration Era
Superagents coordinate specialized agents. Load a playbook in Claude Code and let the mesh deliver.
systemprompt core playbooks show content_blog
# Then in Claude Code: "Run the content_blog playbook and publish the results"
Level 4 is the endgame: agents that regenerate, learn from performance metrics, and orchestrate other agents.
What's Included
See the full documentation at systemprompt.io/documentation for:
- Extensions: Web publishing, memory systems, MCP servers
- Services: Agents, AI providers, authentication, analytics
- Configuration: Profiles, secrets, tenants, deployment
- CLI Reference: Complete command reference
Playbooks
Playbooks are machine-readable instruction guides that eliminate AI hallucination. Instead of letting agents guess CLI syntax (and waste tokens on invented flags), playbooks provide pre-tested, deterministic commands for every operation.
The problem they solve: LLMs frequently invent non-existent CLI commands. Playbooks encode verified commands in JSON format, ensuring agents execute what actually works.
# List all playbooks
systemprompt core playbooks list
# Start here, the master index
systemprompt core playbooks show guide_start
# Show any playbook
systemprompt core playbooks show <playbook_id>
Playbook Categories
| Category | Prefix | Purpose |
|---|---|---|
| Guide | guide_* | Onboarding, meta-documentation, entry points |
| CLI | cli_* | All command-line operations |
| Build | build_* | Development standards and checklists |
| Content | content_* | Content creation workflows |
Essential Playbooks
| Playbook | Description |
|---|---|
guide_start | Master index, read this first |
guide_coding-standards | Rust patterns and code standards |
cli_agents | Agent management commands |
cli_deploy | Deployment procedures |
build_extension-checklist | Building Rust extensions |
content_blog | Blog publishing workflow |
Self-Repair Protocol
When a command fails, agents fix the playbook rather than retry with guesses. This creates a feedback loop where instructions improve over time.
Learn more: systemprompt.io/playbooks
Architecture
┌─────────────────────────────────────────────────────────────────┐
│ Your Project (this template) │
│ │
│ extensions/ ← Your Rust code │
│ ├── web/ Web publishing, themes, SEO │
│ ├── soul/ Long-term memory system │
│ ├── mcp/ MCP servers (3 included) │
│ └── cli/ CLI extensions │
│ │
│ services/ ← Your configuration (YAML/Markdown) │
│ ├── agents/ Agent definitions │
│ ├── skills/ Skill configurations │
│ ├── playbook/ 96 operational playbooks │
│ └── web/ Themes, templates, homepage │
└─────────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────┐
│ systemprompt-core (Cargo dependency) │
│ │
│ Provides: │
│ ├── API server + CLI │
│ ├── Agent runtime + A2A protocol │
│ ├── MCP server hosting │
│ ├── Auth (OAuth2 + WebAuthn) │
│ ├── Memory systems │
│ └── Observability + logging │
└─────────────────────────────────────────────────────────────────┘
Key rules:
- Rust code →
extensions/ - Configuration →
services/ - Core is a Cargo dependency (from crates.io)
Project Structure
systemprompt-template/
├── extensions/ # Your Rust code
│ ├── web/ # Web publishing extension
│ ├── soul/ # Memory system extension
│ ├── mcp/ # MCP servers
│ │ ├── systemprompt/ # CLI execution (admin-only)
│ │ ├── soul/ # Memory operations
│ │ └── content-manager/ # Content creation
│ └── cli/ # CLI extensions
│
├── services/ # Configuration (YAML/Markdown only)
│ ├── agents/ # Agent definitions
│ ├── skills/ # Skill configurations
│ ├── mcp/ # MCP server configs
│ ├── playbook/ # 96 operational playbooks
│ └── web/ # Theme, templates, homepage config
│
├── Cargo.toml # Workspace manifest (systemprompt-core via Cargo)
└── justfile # Development commands
Commands
| Command | Description |
|---|---|
just build | Build the project |
just login | Authenticate with cloud |
just tenant | Create tenant (database, profile, migrations) |
just start | Start all services |
just deploy | Build and deploy to cloud |
just migrate | Run database migrations |
What's Next?
Customize & Develop
Build agents, skills, and MCP servers tailored to your needs.
- Add new agents in
services/agents/ - Define custom skills in
services/skills/ - Build MCP servers in
extensions/mcp/ - See:
systemprompt core playbooks show build_getting-started
Deploy Securely
Ship to production with OAuth, HTTPS, and full audit trails.
just login
just tenant create --region iad
just profile create production
just deploy --profile production
See: systemprompt core playbooks show cli_deploy
Automate & Schedule
Run jobs on schedules and build agentic workflows.
- Cron-based job scheduling
- Event-driven triggers
- Agentic workflow pipelines
See: systemprompt core playbooks show cli_jobs
Built on systemprompt-core
This template extends systemprompt-core, a production-ready AI agent orchestration library that provides:
- Complete runtime: API server, agent runtime, MCP hosting
- Authentication: OAuth2 for APIs, WebAuthn for passwordless login
- Memory systems: Long-term, short-term, and working memory
- A2A protocol: Agent-to-agent communication and orchestration
- Observability: Full request logging and audit trails
Works with Claude Code, Claude Desktop, ChatGPT, and any MCP-compatible tool.
License
MIT, see LICENSE
Depends on systemprompt-core (FSL-1.1-ALv2).
Documentation · Discord · systemprompt-core · Issues
Questions or issues? Join us on Discord for help.