Notion MCP Challenge Entry — Use Notion as the human-in-the-loop command center for managing OpenClaw AI agents.
AgentOps turns your Notion workspace into a fully functional agent operations control plane. Monitor your OpenClaw fleet, dispatch tasks, track token usage, tune agent prompts, and backup your entire configuration — all through Notion.
Humans stay in control. Every agent, task, and configuration lives in Notion. Edit a page to pause an agent. Change a priority by updating a select field. Notion is the database.
Real-time overview of your OpenClaw agent fleet — 18 agents, success rate, token usage, pipeline health, and recent activity.
All 18 OpenClaw agents with status, schedules, and one-click pause/resume. Filter by type: cron, monitor, heartbeat, subagent, or demo agents.
Priority-based task queue with status tracking. Create tasks manually or let the OpenClaw runtime dispatch them automatically.
Full execution history — every agent run with input, output, duration, and token count.
Failure alerts with acknowledge workflow. No more digging through log files.
Per-agent token breakdown, 7-day trends, and top consumers across your OpenClaw fleet.
Push OpenClaw configuration files (SOUL.md, AGENTS.md, TOOLS.md, prompts) to Notion. Secrets auto-redacted.
Bidirectional prompt sync between disk and Notion. Edit prompts in Notion, apply live to running OpenClaw agents.
Full fleet backup to Notion. Export/import agent configurations as portable JSON bundles.
All agent data lives directly in your Notion workspace — Agent Registry, Task Queue, Run Log, Alerts, Backups, and Agent Prompts. No external database.
- 🔗 Notion as Database — No external DB. Notion databases store agents, tasks, runs, and alerts
- 🤖 3 Built-in Agents — Summarizer, Code Reviewer, Sentiment Analyzer (no external AI APIs needed)
- 📋 Task Queue — Priority-based, auto-dispatched to matching agents every 10 seconds
- 📊 Real-time Dashboard — Live metrics: success rate, latency, throughput, fleet health
- 🚨 Alert System — Automatic error alerts with acknowledge workflow
- ⚡ WebSocket Updates — Real-time UI updates when tasks complete
- 🔐 Notion OAuth 2.0 — Secure authentication, no API keys to manage
- 🦀 OpenClaw Fleet Sync — Pull live agent data from
openclaw cron listinto the dashboard - 🏷️ OpenClaw Badges — Visual indicators distinguishing OpenClaw agents from demo agents
- 📅 Schedule Display — Cron schedules shown in human-readable format for each agent
- 💰 Token Dashboard — All-time, daily, weekly token consumption
- 📊 Per-Agent Breakdown — Sortable table with total, average, and last-run token counts
- 📈 7-Day Trend — Visual bar chart of token usage over the past week
- 📤 Push to Notion — Sync SOUL.md, AGENTS.md, TOOLS.md, USER.md, HEARTBEAT.md, IDENTITY.md
- 🔒 Secret Redaction — Automatically strips tokens, passwords, and API keys before sync
- 📝 Markdown to Blocks — Converts markdown to Notion blocks (headings, lists, code, annotations)
- ✏️ Bidirectional Sync — Edit prompts in Notion or on disk, sync either direction
- 🔍 Diff Detection — See which prompts are in sync and which have diverged
- 🚀 Apply Live — Pull from Notion and apply to running OpenClaw agents with one click
- 🗄️ Full Backup — Workspace files, prompts, cron definitions, agent registry to a Notion page
- ♻️ Restore — Write files back to disk from any Notion backup
- 📦 Export Fleet — Download entire agent fleet as a portable JSON bundle
- 📥 Import Fleet — Upload a JSON bundle to recreate agents on a fresh OpenClaw instance
┌─────────────────────────────────────────────────────────────┐
│ NOTION WORKSPACE │
│ │
│ ┌──────────────┐ ┌──────────────┐ ┌─────────┐ ┌────────┐ │
│ │Agent Registry│ │ Task Queue │ │ Run Log │ │ Alerts │ │
│ │ (18 agents) │ │ (78 tasks) │ │(78 runs)│ │(2 errs)│ │
│ └──────┬───────┘ └──────┬───────┘ └────┬────┘ └───┬────┘ │
│ │ Notion API (MCP) │ │ │
└─────────┼────────────────┼──────────────┼───────────┼────────┘
│ │ │ │
┌─────────┴────────────────┴──────────────┴───────────┴────────┐
│ AGENTOPS SERVER │
│ │
│ ┌──────────┐ ┌──────────┐ ┌───────────┐ ┌───────────────┐ │
│ │ OAuth │ │ REST │ │ Runtime │ │ OpenClaw │ │
│ │ Flow │ │ API │ │ (10s poll)│ │ Integration │ │
│ └──────────┘ └──────────┘ └───────────┘ └───────────────┘ │
│ │
│ ┌─────────────────────┐ ┌────────────────────────────────┐ │
│ │ Demo Agents │ │ OpenClaw Fleet (14 cron) │ │
│ │ • Summarizer │ │ • newsletter-scout/writer/pub │ │
│ │ • Code Reviewer │ │ • podcast-generator │ │
│ │ • Sentiment │ │ • job-scout/followup │ │
│ └─────────────────────┘ │ • go-train-monitor │ │
│ │ • daily-backup, heartbeat... │ │
│ └────────────────────────────────┘ │
└───────────────────────────────────────────────────────────────┘
│
┌─────────┴─────────────────────────────────────────────────────┐
│ REACT DASHBOARD │
│ │
│ ┌──────────┐ ┌────────┐ ┌───────┐ ┌────────┐ ┌──────────┐ │
│ │Dashboard │ │ Agents │ │ Tasks │ │ Tokens │ │Workspace │ │
│ ├──────────┤ ├────────┤ ├───────┤ ├────────┤ ├──────────┤ │
│ │ Runs │ │ Alerts │ │Tuning │ │ Backup │ │ Clone │ │
│ └──────────┘ └────────┘ └───────┘ └────────┘ └──────────┘ │
└────────────────────────────────────────────────────────────────┘
- Node.js 18+
- A Notion Integration (public or internal)
- OpenClaw (optional, for fleet integration)
git clone https://github.com/awsdataarchitect/agentops.git
cd agentops
npm run setupcp server/.env.example server/.envEdit server/.env:
NOTION_CLIENT_ID=your_notion_client_id
NOTION_CLIENT_SECRET=your_notion_client_secret
S3_BACKUP_BUCKET=your-backup-bucket
PORT=3000
npm start- Click "Connect to Notion"
- Authorize the integration
- Click "Setup" to create databases and seed agents
- Start creating tasks or explore the dashboard
| Agent | Input | Output | How It Works |
|---|---|---|---|
| Summarizer | URL | Extractive summary | Sentence scoring by word frequency + position |
| Code Reviewer | Code snippet | Lint feedback | 20+ regex rules for security, style, best practices |
| Sentiment Analyzer | Text | Score (-1 to 1) + label | Keyword scoring with negation + intensifiers |
All agents are self-contained — no external AI API keys required.
- Human creates a task in the dashboard or directly in Notion
- Runtime polls the Task Queue every 10 seconds
- Matches task to an active agent by type
- Executes the agent, updating status in real-time
- Writes results to the Run Log in Notion
- Creates alerts if anything fails
- Human reviews in the dashboard or Notion
The key insight: Notion is both the UI and the database. Humans can pause agents, reprioritize tasks, review outputs, and acknowledge alerts — all by editing Notion pages.
agentops/
├── server/
│ ├── src/
│ │ ├── index.js # Express + WebSocket server
│ │ ├── routes/
│ │ │ ├── auth.js # Notion OAuth flow
│ │ │ ├── setup.js # Database creation + seeding
│ │ │ ├── agents.js # Agent CRUD
│ │ │ ├── tasks.js # Task queue
│ │ │ ├── runs.js # Run log
│ │ │ ├── alerts.js # Alert management
│ │ │ ├── dashboard.js # Aggregated metrics
│ │ │ ├── dispatch.js # Manual dispatch
│ │ │ ├── openclaw.js # OpenClaw sync + events
│ │ │ ├── tokens.js # Token analytics
│ │ │ ├── workspace.js # Workspace sync
│ │ │ ├── tuning.js # Agent tuning
│ │ │ └── backup.js # Backup + clone
│ │ ├── services/
│ │ │ ├── notion.js # Notion API + schemas
│ │ │ ├── runtime.js # Agent runtime (poll + dispatch)
│ │ │ ├── state.js # Persistent state
│ │ │ ├── notion-utils.js # Block chunking helpers
│ │ │ ├── openclaw-agents.js # OpenClaw agent definitions
│ │ │ ├── openclaw-sync.js # Live cron sync
│ │ │ ├── workspace-sync.js # File → Notion sync
│ │ │ ├── agent-tuning.js # Bidirectional prompt sync
│ │ │ ├── agent-backup.js # Backup/restore
│ │ │ └── agent-clone.js # Export/import fleet
│ │ ├── agents/
│ │ │ ├── summarizer.js # URL → summary
│ │ │ ├── code-reviewer.js # Code → feedback
│ │ │ └── sentiment.js # Text → sentiment
│ │ └── scripts/
│ │ └── backfill-history.js # Historical data generator
│ └── .env.example
├── client/
│ ├── src/
│ │ ├── App.jsx # Router + layout
│ │ ├── api.js # API client
│ │ ├── components/
│ │ │ ├── Sidebar.jsx # Navigation
│ │ │ ├── Dashboard.jsx # Metrics + fleet overview
│ │ │ ├── AgentRegistry.jsx # Agent cards + controls
│ │ │ ├── TaskQueue.jsx # Task table + create
│ │ │ ├── RunLog.jsx # Execution history
│ │ │ └── Alerts.jsx # Alert notifications
│ │ └── pages/
│ │ ├── Tokens.jsx # Token analytics
│ │ ├── Workspace.jsx # Workspace sync
│ │ ├── Tuning.jsx # Agent tuning
│ │ └── Backup.jsx # Backup & clone
│ └── vite.config.js
├── screenshots/ # App screenshots
└── README.md
- Go to notion.so/my-integrations
- Create a new integration
- Set redirect URI to
http://localhost:3000/auth/callback - Copy Client ID and Client Secret to
.env - Enable capabilities: Read content, Update content, Insert content
Most agent frameworks run autonomously with minimal oversight. AgentOps flips this by making Notion the control plane for OpenClaw:
- Notion is the source of truth. Not a dashboard that syncs — the actual database.
- Humans edit Notion pages to control agents. Pause, reprioritize, configure.
- Full audit trail in Run Log. Every execution recorded.
- Alert workflow ensures humans are notified and can acknowledge.
- Workspace sync lets you tune agent prompts in Notion's rich editor.
- Fleet backup means your entire OpenClaw configuration is portable and restorable.
This is what "human-in-the-loop" should look like: a full operational command center where humans have complete visibility and control over their AI agent fleet.
MIT
Built for the Notion MCP Challenge | Powered by OpenClaw