agt0 for Openclaw

agt0 is a unified local-first storage solution for AI agents that combines a SQLite database, a virtual filesystem, and long-term memory into a single .db file.

liuhao6741
v1.0.1
Mar 21, 2026
0
671
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install agt0

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install agt0 using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).

Prefer to download?

Get the raw skill files in a ZIP archive.

What is agt0?

agt0 is a robust storage skill designed for developers building AI agents that require persistent state, structured data handling, and file management without the overhead of external cloud services. By leveraging SQLite, it allows agents to treat a single local file as both a relational database and a virtual filesystem, ensuring all data remains local-first and private.

This skill is a vital addition to the Openclaw Skills library, providing agents with the ability to query CSV, TSV, and JSONL files directly using SQL. It eliminates the need for complex data ingestion pipelines by providing virtual table modules that map flat files into queryable relational structures in real-time.

agt0 Use Cases

  • Persisting AI agent memory and session states across restarts.
  • Querying local log files (JSONL) or datasets (CSV/TSV) using standard SQL syntax.
  • Managing a virtual filesystem inside a single database file for portable agent environments.
  • Creating branched development environments for agents to test logic without affecting production data.
  • Zero-import data analysis where agents can "read" files as if they were native database tables.

How agt0 Works

  1. The user initializes a local database file using the agt0 CLI, which creates the underlying SQLite structure.
  2. Agents use the virtual filesystem (VFS) layer to read, write, or append files via SQL functions like fs_read() or fs_write().
  3. Structured files such as CSVs are accessed through Table-Valued Functions (TVFs) that parse data into JSON objects on the fly.
  4. For high-performance queries, virtual tables are created using the expand modules to map file headers to native SQL columns.
  5. All operations are performed locally, with the agent interacting through either the CLI, Node.js API, or direct SQL commands.

agt0 Setup

Install the agt0 package globally and initialize your first database to begin using Openclaw Skills for storage:

npm install -g @seekcontext/agt0
agt0 init my_agent_db
agt0 use my_agent_db

agt0 Data Schema & Taxonomy

agt0 organizes all data within a single .db file. It uses an internal _fs table to manage the virtual filesystem metadata. Users can define their own relational tables alongside this virtual file storage.

Component Implementation Description
Relational Data Standard SQLite Tables User-defined schemas for structured data.
Virtual Filesystem _fs (Internal Table) Stores file paths, blobs, and metadata.
Flat Files Virtual Tables CSV, TSV, and JSONL files mapped via TVFs.
Metadata ISO 8601 Strings Timestamps for file modification (mtime).

agt0 Advanced Features

  • Advanced path globbing support (e.g., /logs/**/*.log) for multi-file querying.
  • Automatic SQL rewriting in the CLI to transform file paths into queryable virtual tables.
  • Database branching to instantly clone agent states for debugging or staging.
  • Low-level byte manipulation with fs_read_at and fs_write_at for binary data handling.
  • Environment-based constraints to limit file sizes, row counts, and memory usage for safety.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*