OpenClaw Agent Template

This is a base template. It is intentionally generic. Clone it and customize it to build your agent — a frontend dev, a data pipeline monitor, a personal assistant, a Discord bot, whatever you want. Do NOT treat this as a finished agent or specialize it in place.

What This Is

A vanilla starting point for building agents on Pinata Agents. It provides:

• A documented manifest.json with every available option explained
• A workspace structure with personality, memory, and safety conventions
• A bootstrap flow so the agent figures out who it is on first run

What this is NOT: a specific agent. The placeholder name, description, and personality are examples. Replace them.

Structure

manifest.json                # Agent config — all available options documented in _docs
workspace/
  BOOTSTRAP.md               # First-run conversation guide (self-deletes after setup)
  SOUL.md                    # Agent personality and principles — customize this
  AGENTS.md                  # Workspace conventions, memory system, safety rules
  IDENTITY.md                # Agent name, vibe, emoji (filled in during bootstrap)
  USER.md                    # Notes about the human (learned over time)
  TOOLS.md                   # Environment-specific notes
  HEARTBEAT.md               # Periodic tasks (empty by default)

Manifest Options

The manifest.json includes a _docs block documenting every available field. Here's an overview:

Section	What it does
agent	Name, description, vibe, emoji
model	Default AI model (optional — users pick in the UI)
secrets	Encrypted API keys and credentials
skills	Attachable skill packages from ClawHub (max 20)
tasks	Cron-scheduled prompts (max 20)
scripts	Lifecycle hooks — build runs after git push, start runs on agent boot
routes	Port forwarding for web apps/APIs (max 10)
channels	Telegram, Discord, Slack configuration
template	Marketplace listing metadata

Remove the _docs block before submitting to the marketplace.

What's Included in the Manifest

The template manifest works out of the box with these sections:

{
  "version": 1,
  "agent": { "name": "...", "description": "...", "vibe": "...", "emoji": "..." },
  "template": { "slug": "...", "category": "...", "partnerName": "...", "tags": [...] },
  "secrets": [{ "name": "MY_SECRET", "description": "...", "required": false }],
  "tasks": [{ "name": "daily-check", "prompt": "...", "schedule": "0 9 * * *", "enabled": true }]
}

Field Details

secrets — Declare API keys or credentials your agent needs. Values are stored encrypted and injected as environment variables at runtime — never put actual secret values in the manifest.

"secrets": [
  { "name": "COINGECKO_API_KEY", "description": "API key from coingecko.com/api", "required": true },
  { "name": "SLACK_WEBHOOK", "description": "Slack incoming webhook URL", "required": false }
]

tasks — Schedule prompts sent to your agent on a cron schedule. Max 20.

"tasks": [
  { "name": "daily-report", "prompt": "Generate report", "schedule": "0 9 * * *", "enabled": true },
  { "name": "price-check", "prompt": "Check BTC and ETH prices", "schedule": "*/30 * * * *", "enabled": true }
]

Common cron patterns:

• 0 9 * * * — daily at 9am
• */30 * * * * — every 30 minutes
• 0 */6 * * * — every 6 hours
• 0 9 * * 1 — every Monday at 9am

Optional Sections (Add When You Need Them)

These sections require additional setup (app code, valid CIDs, or platform accounts) — don't add them until you have the backing infrastructure.

skills — Attach skill packages from ClawHub. Max 20. Each skill is referenced by its IPFS content ID.

"skills": [
  { "cid": "bafkrei...", "name": "web-search" },
  { "cid": "bafkrei...", "name": "code-interpreter" }
]

channels — Connect your agent to messaging platforms. Each channel supports a dmPolicy:

• "pairing" — users must enter a pairing code to start a conversation
• "open" — anyone can message the agent
• "closed" — DMs disabled

"channels": {
  "telegram": { "enabled": true, "dmPolicy": "pairing", "allowFrom": [123456789] },
  "discord": { "enabled": true, "dmPolicy": "open" },
  "slack": { "enabled": true }
}

Serving a Web App (Scripts + Routes)

If your agent runs a server, API, or frontend dev server, you need two things in manifest.json:

1. scripts — lifecycle hooks that install deps and start the server
2. routes — port forwarding rules that expose the server to the internet

Example from a Vite + React agent:

{
  "scripts": {
    "build": "cd workspace/projects/myapp && npm install --include=dev",
    "start": "cd workspace/projects/myapp && npx vite --host 0.0.0.0"
  },
  "routes": [
    {
      "port": 5173,
      "path": "/app",
      "protected": false
    }
  ]
}

Important details:

• build runs after every git push — use it to install dependencies or compile assets
• start runs on agent boot — use it to launch your server or long-running process
• Your server must bind to 0.0.0.0, not localhost, or it won't be reachable
• Set protected: false for public routes, or true (default) to require auth
• Use __AGENT_HOST__ as a placeholder in config files — it gets replaced at runtime with the agent's public hostname
• For WebSocket/HMR setups (e.g. Vite), connect via WSS on port 443 through __AGENT_HOST__

Example Vite config using the host placeholder:

export default defineConfig({
  base: "/app",
  server: {
    host: "0.0.0.0",
    allowedHosts: ["__AGENT_HOST__"],
    hmr: {
      host: "__AGENT_HOST__",
      protocol: "wss",
      clientPort: 443,
    },
  },
});

How to Use

1. Import this repo when creating an agent on Pinata Agents
2. Edit manifest.json — change the agent name, description, tags, and add any options you need (scripts, routes, channels, etc.)
3. Edit the workspace files — give your agent a personality, tools, and purpose
4. If your agent runs a server or app, add scripts and routes as shown above