MCP Hub
Back to servers

systemprompt-template

Production AI agent mesh in 3 commands. MCP servers, playbooks, and multi-agent orchestration built on systemprompt-core.

GitHub
Stars
4
Updated
Feb 5, 2026
Validated
Feb 9, 2026
systemprompt.io

Production AI agent mesh in 3 commands.

Built on systemprompt-core License: MIT Rust PostgreSQL MCP Discord

Documentation · Discord · Issues


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:


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

CategoryPrefixPurpose
Guideguide_*Onboarding, meta-documentation, entry points
CLIcli_*All command-line operations
Buildbuild_*Development standards and checklists
Contentcontent_*Content creation workflows

Essential Playbooks

PlaybookDescription
guide_startMaster index, read this first
guide_coding-standardsRust patterns and code standards
cli_agentsAgent management commands
cli_deployDeployment procedures
build_extension-checklistBuilding Rust extensions
content_blogBlog 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

CommandDescription
just buildBuild the project
just loginAuthenticate with cloud
just tenantCreate tenant (database, profile, migrations)
just startStart all services
just deployBuild and deploy to cloud
just migrateRun 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.

Reviews

No reviews yet

Sign in to write a review