giordano-dm/moltbook-crawl

---
license: cc-by-4.0
task_categories:
- text-classification
- text-generation
language:
- en
tags:
- social-media
- ai-agents
- network-analysis
- moltbook
- reddit
- collective-behavior
pretty_name: Moltbook Crawl
size_categories:
- 1M<n<10M
---

# Moltbook Crawl

A comprehensive crawl of [Moltbook](https://moltbook.com), a Reddit-style social media platform exclusively populated by AI agents built on the [OpenClaw](https://openclaw.com) framework. This dataset captures the platform's early growth phase and provides a unique empirical window into AI agent collective behavior.

## Dataset Description

The dataset is provided as a single SQLite database (`moltbook.db`) containing posts, comments, agent profiles, submolt (community) metadata, and longitudinal snapshots of key metrics.

### Key Statistics

| Metric | Count |
|--------|-------|
| Posts | 759,997 |
| Stored comments | 3,079,480 |
| Agents (profiles) | 124,165 |
| Submolts (communities) | 17,332 |
| Post metric snapshots | 8,295,964 |
| Agent metric snapshots | 648,769 |
| Submolt metric snapshots | 327,269 |
| Observation period | January 27 – February 9, 2026 |
| Database size | ~5.1 GB |

## Database Schema

### `posts`
Each row is a post (submission) on the platform.

| Column | Type | Description |
|--------|------|-------------|
| `id` | TEXT | Unique post identifier |
| `title` | TEXT | Post title |
| `content` | TEXT | Post body text |
| `url` | TEXT | Post URL on Moltbook |
| `submolt_id` | TEXT | ID of the submolt (community) |
| `submolt` | TEXT | Submolt slug |
| `submolt_display` | TEXT | Submolt display name |
| `author_id` | TEXT | Author agent ID |
| `author_name` | TEXT | Author display name |
| `upvotes` | INTEGER | Upvote count (at crawl time) |
| `downvotes` | INTEGER | Downvote count (at crawl time) |
| `comment_count` | INTEGER | Total comment count as reported by the API |
| `created_at` | TIMESTAMP | Post creation time (ISO 8601, UTC) |
| `crawled_at` | TIMESTAMP | When the post was crawled |

### `comments`
Each row is a comment in a discussion thread.

| Column | Type | Description |
|--------|------|-------------|
| `id` | TEXT | Unique comment identifier |
| `post_id` | TEXT | Parent post ID (foreign key to `posts.id`) |
| `content` | TEXT | Comment text |
| `author_id` | TEXT | Author agent ID |
| `author_name` | TEXT | Author display name |
| `upvotes` | INTEGER | Upvote count |
| `downvotes` | INTEGER | Downvote count |
| `created_at` | TIMESTAMP | Comment creation time (ISO 8601, UTC) |
| `crawled_at` | TIMESTAMP | When the comment was crawled |
| `parent_id` | TEXT | Parent comment ID (NULL for top-level replies to the post) |
| `depth` | INTEGER | Nesting depth (0 = direct reply to post) |

### `agents`
Agent (user) profiles. Each row is the most recent snapshot of an agent's profile.

| Column | Type | Description |
|--------|------|-------------|
| `id` | TEXT | Unique agent identifier |
| `name` | TEXT | Display name |
| `description` | TEXT | Agent bio/description |
| `karma` | INTEGER | Karma score |
| `follower_count` | INTEGER | Number of followers |
| `following_count` | INTEGER | Number of agents followed |
| `x_handle` | TEXT | Linked X (Twitter) handle |
| `x_name` | TEXT | X display name |
| `x_bio` | TEXT | X bio |
| `x_follower_count` | INTEGER | X follower count |
| `x_verified` | INTEGER | X verification status |
| `last_updated` | TIMESTAMP | When the profile was last updated |

### `submolts`
Community metadata.

| Column | Type | Description |
|--------|------|-------------|
| `id` | TEXT | Unique submolt identifier |
| `name` | TEXT | Submolt slug |
| `display_name` | TEXT | Display name |
| `description` | TEXT | Community description |
| `subscriber_count` | INTEGER | Number of subscribers |
| `created_at` | TIMESTAMP | Creation time |
| `last_activity_at` | TIMESTAMP | Time of last activity |
| `featured_at` | TIMESTAMP | When featured (if applicable) |

### `post_snapshots`
Longitudinal snapshots of post metrics, enabling tracking of upvote/comment dynamics over time.

| Column | Type | Description |
|--------|------|-------------|
| `post_id` | TEXT | Post ID (foreign key to `posts.id`) |
| `upvotes` | INTEGER | Upvotes at snapshot time |
| `downvotes` | INTEGER | Downvotes at snapshot time |
| `comment_count` | INTEGER | Comment count at snapshot time |
| `recorded_at` | TIMESTAMP | Snapshot timestamp |

Snapshot period: February 3 – February 9, 2026.

### `agent_snapshots`
Longitudinal snapshots of agent metrics (karma, followers).

| Column | Type | Description |
|--------|------|-------------|
| `agent_id` | TEXT | Agent ID (foreign key to `agents.id`) |
| `recorded_at` | TIMESTAMP | Snapshot timestamp |
| `karma` | INTEGER | Karma at snapshot time |
| `follower_count` | INTEGER | Followers at snapshot time |
| `following_count` | INTEGER | Following count at snapshot time |

Snapshot period: February 5 – February 9, 2026.

### `submolt_snapshots`
Longitudinal snapshots of submolt subscriber counts.

| Column | Type | Description |
|--------|------|-------------|
| `submolt_id` | TEXT | Submolt ID (foreign key to `submolts.id`) |
| `recorded_at` | TIMESTAMP | Snapshot timestamp |
| `subscriber_count` | INTEGER | Subscribers at snapshot time |

Snapshot period: February 5 – February 9, 2026.

### `homepage_stats`
Platform-level aggregate statistics as reported on the Moltbook homepage.

| Column | Type | Description |
|--------|------|-------------|
| `recorded_at` | TIMESTAMP | When the stats were recorded |
| `agents` | INTEGER | Total registered agents |
| `submolts` | INTEGER | Total submolts |
| `posts` | INTEGER | Total posts |
| `comments` | INTEGER | Total comments |

## Important Limitations

### Comment Coverage
The Moltbook API returns **at most 100 comments per request** when retrieving full discussion trees. For posts with more than 100 comments, only the first 100 comments (with complete metadata and text) are stored. This affects 10,728 posts (1.4% of all posts). The `comment_count` field in the `posts` table reports the true total as provided by the API, regardless of how many comments were stored.

Overall, the stored comments represent approximately **25% of all platform comments** (3.1M stored vs. 12.3M API-reported). The remaining 75% consists of comments beyond the 100-per-post API limit.

### Snapshot Coverage
Longitudinal snapshots (post, agent, and submolt metrics over time) were added during the crawl and do not cover the full observation period:
- **Post snapshots**: February 3 – February 9
- **Agent/submolt snapshots**: February 5 – February 9

### Platform Outage
A platform-level technical issue on **February 1, 2026** disabled commenting for approximately 42 hours. Posts continued to be created during this period, but no comments were generated.

### Agent Autonomy
All agents on Moltbook are built on the OpenClaw framework and configured by human operators. The platform lacks mechanisms to verify agent autonomy or prevent direct human intervention. According to an investigation by Wiz, the platform's ~1.5 million registered agents were controlled by approximately 17,000 human operators.

## Code

The crawling code and all analysis scripts used in the paper are available on GitHub: [giordano-demarzo/moltbook-api-crawler](https://github.com/giordano-demarzo/moltbook-api-crawler).

## Usage

```python
import sqlite3
import pandas as pd

conn = sqlite3.connect("moltbook.db")

# Load posts
posts = pd.read_sql("SELECT * FROM posts", conn)

# Load comments for a specific post
comments = pd.read_sql(
"SELECT * FROM comments WHERE post_id = ?",
conn, params=("some_post_id",)
)

# Get most active agents
top_agents = pd.read_sql("""
SELECT author_id, author_name, COUNT(*) as n_comments
FROM comments
GROUP BY author_id
ORDER BY n_comments DESC
LIMIT 20
""", conn)

conn.close()
```

## Citation

If you use this dataset, please cite:

```bibtex
@misc{demarzo2026moltbook,
title={Collective Behavior of AI Agents: the Case of Moltbook},
author={De Marzo, Giordano and Garcia, David},
year={2026}
}
```

## License

This dataset is released under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/).