open-agent-db 2.0.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- open_agent_db-2.0.0/LICENSE +21 -0
- open_agent_db-2.0.0/PKG-INFO +256 -0
- open_agent_db-2.0.0/README.md +227 -0
- open_agent_db-2.0.0/cli/__init__.py +2 -0
- open_agent_db-2.0.0/cli/db.py +601 -0
- open_agent_db-2.0.0/cli/installer.py +131 -0
- open_agent_db-2.0.0/cli/main.py +272 -0
- open_agent_db-2.0.0/cli/vector_engine.py +184 -0
- open_agent_db-2.0.0/open_agent_db.egg-info/PKG-INFO +256 -0
- open_agent_db-2.0.0/open_agent_db.egg-info/SOURCES.txt +21 -0
- open_agent_db-2.0.0/open_agent_db.egg-info/dependency_links.txt +1 -0
- open_agent_db-2.0.0/open_agent_db.egg-info/entry_points.txt +4 -0
- open_agent_db-2.0.0/open_agent_db.egg-info/requires.txt +1 -0
- open_agent_db-2.0.0/open_agent_db.egg-info/top_level.txt +1 -0
- open_agent_db-2.0.0/pyproject.toml +59 -0
- open_agent_db-2.0.0/setup.cfg +4 -0
- open_agent_db-2.0.0/tests/test_occupations.py +53 -0
- open_agent_db-2.0.0/web/app.js +923 -0
- open_agent_db-2.0.0/web/data/catalog_index.json +1 -0
- open_agent_db-2.0.0/web/data/stats.json +458 -0
- open_agent_db-2.0.0/web/index.html +350 -0
- open_agent_db-2.0.0/web/serve.py +127 -0
- open_agent_db-2.0.0/web/style.css +136 -0
|
@@ -0,0 +1,21 @@
|
|
|
1
|
+
MIT License
|
|
2
|
+
|
|
3
|
+
Copyright (c) 2026 Open-Agent-DB Contributors
|
|
4
|
+
|
|
5
|
+
Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
6
|
+
of this software and associated documentation files (the "Software"), to deal
|
|
7
|
+
in the Software without restriction, including without limitation the rights
|
|
8
|
+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
9
|
+
copies of the Software, and to permit persons to whom the Software is
|
|
10
|
+
furnished to do so, subject to the following conditions:
|
|
11
|
+
|
|
12
|
+
The above copyright notice and this permission notice shall be included in all
|
|
13
|
+
copies or substantial portions of the Software.
|
|
14
|
+
|
|
15
|
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
16
|
+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
17
|
+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
18
|
+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
19
|
+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
20
|
+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
|
21
|
+
SOFTWARE.
|
|
@@ -0,0 +1,256 @@
|
|
|
1
|
+
Metadata-Version: 2.4
|
|
2
|
+
Name: open-agent-db
|
|
3
|
+
Version: 2.0.0
|
|
4
|
+
Summary: The universal open catalog and package manager for AI Agent Skills & Model Context Protocol (MCP) Servers
|
|
5
|
+
Author: Enes Öztürk
|
|
6
|
+
License: MIT
|
|
7
|
+
Project-URL: Homepage, https://nsozturk.github.io/open-agent-db/
|
|
8
|
+
Project-URL: Repository, https://github.com/nsozturk/open-agent-db
|
|
9
|
+
Project-URL: Documentation, https://nsozturk.github.io/open-agent-db/
|
|
10
|
+
Project-URL: Hugging Face, https://huggingface.co/datasets/ns0bj/open-agent-db
|
|
11
|
+
Project-URL: Issues, https://github.com/nsozturk/open-agent-db/issues
|
|
12
|
+
Keywords: ai-agents,mcp,mcp-server,model-context-protocol,claude-code,cursor-rules,vector-search,sqlite,semantic-search,faiss
|
|
13
|
+
Classifier: Development Status :: 4 - Beta
|
|
14
|
+
Classifier: Intended Audience :: Developers
|
|
15
|
+
Classifier: License :: OSI Approved :: MIT License
|
|
16
|
+
Classifier: Programming Language :: Python :: 3
|
|
17
|
+
Classifier: Programming Language :: Python :: 3.8
|
|
18
|
+
Classifier: Programming Language :: Python :: 3.9
|
|
19
|
+
Classifier: Programming Language :: Python :: 3.10
|
|
20
|
+
Classifier: Programming Language :: Python :: 3.11
|
|
21
|
+
Classifier: Programming Language :: Python :: 3.12
|
|
22
|
+
Classifier: Topic :: Software Development :: Libraries
|
|
23
|
+
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
|
|
24
|
+
Requires-Python: >=3.8
|
|
25
|
+
Description-Content-Type: text/markdown
|
|
26
|
+
License-File: LICENSE
|
|
27
|
+
Requires-Dist: rich>=13.0.0
|
|
28
|
+
Dynamic: license-file
|
|
29
|
+
|
|
30
|
+
<p align="center">
|
|
31
|
+
<img src="https://raw.githubusercontent.com/nsozturk/open-agent-db/main/assets/banner.png" alt="Open-Agent-DB — The Universal AI Agent & MCP Hub" width="100%">
|
|
32
|
+
</p>
|
|
33
|
+
|
|
34
|
+
<div align="center">
|
|
35
|
+
|
|
36
|
+
<p>
|
|
37
|
+
<a href="https://www.npmjs.com/package/open-agent-db"><img src="https://img.shields.io/npm/v/open-agent-db?style=flat-square&color=cb3837&logo=npm" alt="NPM Version" /></a>
|
|
38
|
+
<img src="https://img.shields.io/badge/python-3.8%2B-blue?style=flat-square&logo=python&logoColor=white" alt="Python" />
|
|
39
|
+
<img src="https://img.shields.io/badge/node-%3E%3D18.0-339933?style=flat-square&logo=nodedotjs&logoColor=white" alt="Node" />
|
|
40
|
+
<img src="https://img.shields.io/badge/total%20assets-3.48M%2B-purple?style=flat-square" alt="Total Assets" />
|
|
41
|
+
<img src="https://img.shields.io/badge/mcp%20servers-113K%2B-orange?style=flat-square" alt="MCP Servers" />
|
|
42
|
+
<img src="https://img.shields.io/badge/vector%20db-Parquet%20Dense%20Embeddings-success?style=flat-square" alt="Vector DB" />
|
|
43
|
+
<img src="https://img.shields.io/badge/database-SQLite%20FTS5-blue?style=flat-square&logo=sqlite&logoColor=white" alt="Database" />
|
|
44
|
+
<a href="https://huggingface.co/datasets/ns0bj/open-agent-db"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Datasets%20%26%20Vectors-yellow?style=flat-square" alt="Hugging Face" /></a>
|
|
45
|
+
<a href="https://evren.ssyz.org.tr/datasets/u4be40c9fa813/open-agent-db"><img src="https://img.shields.io/badge/EVREN-NLP%20Dataset-14b8a6?style=flat-square" alt="EVREN dataset" /></a>
|
|
46
|
+
<a href="AGENTS.md"><img src="https://img.shields.io/badge/AI%20Agent-Native%20Ready-blueviolet?style=flat-square" alt="AI Agent Ready" /></a>
|
|
47
|
+
<a href="https://nsozturk.github.io/open-agent-db/"><img src="https://img.shields.io/badge/docs-GitHub%20Pages-00bcd4?style=flat-square" alt="Docs" /></a>
|
|
48
|
+
<img src="https://img.shields.io/badge/license-MIT-green?style=flat-square" alt="License" />
|
|
49
|
+
</p>
|
|
50
|
+
|
|
51
|
+
<p align="center">
|
|
52
|
+
<code>ai-agents</code> •
|
|
53
|
+
<code>model-context-protocol</code> •
|
|
54
|
+
<code>mcp-servers</code> •
|
|
55
|
+
<code>claude-skills</code> •
|
|
56
|
+
<code>cursor-rules</code> •
|
|
57
|
+
<code>vector-database</code> •
|
|
58
|
+
<code>semantic-search</code> •
|
|
59
|
+
<code>sqlite-fts5</code> •
|
|
60
|
+
<code>faiss</code>
|
|
61
|
+
</p>
|
|
62
|
+
|
|
63
|
+
**The Universal Open Database and Semantic Vector Search Engine for AI Agent Skills, MCP Servers, Project Directives, and Rules.**
|
|
64
|
+
|
|
65
|
+
[Documentation](https://nsozturk.github.io/open-agent-db/) • [Features](#-features) • [Ecosystems](#-supported-ecosystems) • [Quick Start](#-quick-start) • [Vector Search](#-semantic-vector-search) • [Hugging Face](#-hugging-face-integration) • [License](#-license)
|
|
66
|
+
|
|
67
|
+
</div>
|
|
68
|
+
|
|
69
|
+
---
|
|
70
|
+
|
|
71
|
+
# Open-Agent-DB — Open Dataset & Search Engine for AI Agent Skills and MCP Servers
|
|
72
|
+
|
|
73
|
+
**Open-Agent-DB is a free, open-source dataset and offline search engine for AI agent skills, MCP servers and coding-agent rules: 3.48M+ assets from Claude Code, Cursor, Codex, Gemini, Copilot and OpenCode in SQLite FTS5 and Parquet vectors.**
|
|
74
|
+
|
|
75
|
+
📚 [Docs](https://nsozturk.github.io/open-agent-db/) · 🔌 [MCP server database](https://nsozturk.github.io/open-agent-db/mcp-servers.html) · 🧠 [Claude Code skills](https://nsozturk.github.io/open-agent-db/claude-code-skills.html) · 🎯 [Cursor rules](https://nsozturk.github.io/open-agent-db/cursor-rules.html) · 🤗 [Dataset on Hugging Face](https://huggingface.co/datasets/ns0bj/open-agent-db)
|
|
76
|
+
|
|
77
|
+
## 🌟 What is Open-Agent-DB?
|
|
78
|
+
|
|
79
|
+
**Open-Agent-DB** is the comprehensive, offline-first open catalog and search engine for the AI Agent era. It unifies **3,480,000+ Agent Assets** and **113,000+ MCP (Model Context Protocol) Servers & Tools** across every major ecosystem:
|
|
80
|
+
|
|
81
|
+
* **Universal Agent Skills**: Canonical `SKILL.md` definitions, multi-step actions, and tool integration scripts.
|
|
82
|
+
* **Project System Rules**: `CLAUDE.md`, `AGENTS.md`, `GEMINI.md`, and `copilot-instructions.md`.
|
|
83
|
+
* **Agent Briefs & Personas**: Anthropic Claude Code (`.claude/agents/*.md`), OpenAI Codex (`.codex/agents/*.md`).
|
|
84
|
+
* **Context & Trigger Rules**: Cursor IDE rules (`.cursorrules`, `.cursor/rules/*.mdc`).
|
|
85
|
+
* **MCP Workflows & Servers**: Official Anthropic MCP Registry, Glama, Smithery, OpenCode (`opencode.json`, `mcp.json`).
|
|
86
|
+
* **Curated System Prompts**: High-impact operational templates and task instructions.
|
|
87
|
+
|
|
88
|
+
Every asset is indexed with author metadata, star ratings, and install commands, and stored with full source code inside SQLite FTS5 databases and dense **vector embeddings**.
|
|
89
|
+
|
|
90
|
+
---
|
|
91
|
+
|
|
92
|
+
## ⚡ Features
|
|
93
|
+
|
|
94
|
+
* 🔍 **Sub-second Full-Text Search (FTS5)**: Query 3.48M+ assets instantly via CLI or web.
|
|
95
|
+
* 🧠 **Semantic Vector Search**: Natural language similarity search using dense vector embeddings over skills, tools, and directives.
|
|
96
|
+
* ⚡ **Zero-Install `npx` Run**: Try instantly without downloading or configuring anything.
|
|
97
|
+
* 🚀 **1-Click Agent Installer**:
|
|
98
|
+
* Automatically installs skills directly to `~/.claude/skills/`, `~/.codex/skills/`, or `.cursor/rules/`.
|
|
99
|
+
* Generates ready-to-use JSON config for Claude Desktop (`claude_desktop_config.json`) and Cursor.
|
|
100
|
+
* 🌐 **Dual-Mode Web Explorer**:
|
|
101
|
+
* **GitHub Pages Static Mode**: Zero backend required! Runs 100% in the browser using a client-side curated index.
|
|
102
|
+
* **Local Mode**: Runs against the full SQLite database for unlimited deep searching.
|
|
103
|
+
* 📦 **Hugging Face Hub Native**: One-command pipeline to export metadata or dense vector embeddings directly into Hugging Face Datasets (`Parquet` / `Arrow`).
|
|
104
|
+
|
|
105
|
+
---
|
|
106
|
+
|
|
107
|
+
## 🧩 Supported Ecosystems
|
|
108
|
+
|
|
109
|
+
| Ecosystem | Primary Target Files | Asset Kinds | Description |
|
|
110
|
+
| :--- | :--- | :--- | :--- |
|
|
111
|
+
| **Universal** | `SKILL.md`, `AGENTS.md`, `AGENT.md` | `skill`, `project_rules` | Cross-platform agent rules and tool skills |
|
|
112
|
+
| **Claude Code** | `CLAUDE.md`, `.claude/agents/*.md`, `.claude/skills/*` | `project_rules`, `agent_brief`, `skill` | Anthropic Claude Code CLI configurations & personas |
|
|
113
|
+
| **Cursor** | `.cursorrules`, `.cursor/rules/*.mdc` | `project_rules`, `trigger_rule` | Cursor IDE rules and trigger-based `.mdc` specs |
|
|
114
|
+
| **OpenAI Codex** | `.codex/agents/*.md`, `.codex-plugin/*`, `mcp.json` | `agent_brief`, `workflow`, `project_rules` | OpenAI Codex CLI agents and plugin specifications |
|
|
115
|
+
| **Google Gemini** | `GEMINI.md`, `.gemini/config/skills/*` | `project_rules`, `skill` | Google Antigravity & Gemini CLI directives |
|
|
116
|
+
| **GitHub Copilot**| `copilot-instructions.md`, `.github/copilot-*` | `project_rules` | GitHub Copilot Workspace custom guidelines |
|
|
117
|
+
| **OpenCode** | `opencode.json`, `.opencode/*` | `workflow`, `skill` | OpenCode agent workflows and tool configs |
|
|
118
|
+
|
|
119
|
+
---
|
|
120
|
+
|
|
121
|
+
## 🚀 Quick Start
|
|
122
|
+
|
|
123
|
+
### 1. Instant Run (Zero Installation via `npx`)
|
|
124
|
+
|
|
125
|
+
You can run Open-Agent-DB immediately using Node.js without installing anything:
|
|
126
|
+
|
|
127
|
+
```bash
|
|
128
|
+
# Search across all indexed assets
|
|
129
|
+
npx open-agent-db search "postgres"
|
|
130
|
+
|
|
131
|
+
# Inspect MCP server details & install command
|
|
132
|
+
npx open-agent-db info "Prisma Postgres"
|
|
133
|
+
|
|
134
|
+
# Generate Claude Desktop MCP JSON config snippet
|
|
135
|
+
npx open-agent-db install "Prisma Postgres"
|
|
136
|
+
```
|
|
137
|
+
|
|
138
|
+
### 2. Global Installation (NPM or Pip)
|
|
139
|
+
|
|
140
|
+
#### Via NPM (Node.js ecosystem):
|
|
141
|
+
```bash
|
|
142
|
+
npm install -g open-agent-db
|
|
143
|
+
open-agent search "docker"
|
|
144
|
+
```
|
|
145
|
+
|
|
146
|
+
#### Via Pip (Python & AI ecosystem):
|
|
147
|
+
```bash
|
|
148
|
+
pip install open-agent-db
|
|
149
|
+
open-agent search "react"
|
|
150
|
+
```
|
|
151
|
+
|
|
152
|
+
---
|
|
153
|
+
|
|
154
|
+
## 🧠 Semantic Vector Search
|
|
155
|
+
|
|
156
|
+
Open-Agent-DB includes a native Vector Search Engine that allows natural language semantic queries:
|
|
157
|
+
|
|
158
|
+
```bash
|
|
159
|
+
# Natural language semantic query
|
|
160
|
+
python3 vector_search.py "automate database schema migrations with safety checks"
|
|
161
|
+
|
|
162
|
+
# Search for specific tools with top-K limit
|
|
163
|
+
python3 vector_search.py "ios swift metal shader visual effects" --top-k 5
|
|
164
|
+
|
|
165
|
+
# Rebuild local vector embeddings index from SQLite
|
|
166
|
+
python3 vector_search.py --build --limit 50000
|
|
167
|
+
```
|
|
168
|
+
|
|
169
|
+
---
|
|
170
|
+
|
|
171
|
+
## 🤗 Hugging Face Integration
|
|
172
|
+
|
|
173
|
+
Open-Agent-DB provides a unified Hub on Hugging Face Datasets ([`ns0bj/open-agent-db`](https://huggingface.co/datasets/ns0bj/open-agent-db)) hosting two core databases:
|
|
174
|
+
|
|
175
|
+
The metadata catalogue (3.65M unified rows + 150K MCP registry items, metadata only, secrets redacted) is also published on Türkiye's EVREN AI platform as an NLP dataset: [`open-agent-db` on EVREN](https://evren.ssyz.org.tr/datasets/u4be40c9fa813/open-agent-db) — uploaded with [evren-platform-api](https://github.com/nsozturk/evren-platform-api).
|
|
176
|
+
|
|
177
|
+
### 1. 🧠 Semantic Vector DB (`open_agent_db_vectors.parquet`)
|
|
178
|
+
Pre-computed dense 128-dimensional float32 vector embeddings over Agent Skills, MCP Servers, and Directives. Ready for instant semantic similarity search via FAISS, cosine distance, or LanceDB.
|
|
179
|
+
|
|
180
|
+
```python
|
|
181
|
+
from datasets import load_dataset
|
|
182
|
+
|
|
183
|
+
# Load vector embeddings directly from Hugging Face
|
|
184
|
+
dataset = load_dataset("ns0bj/open-agent-db", data_files="open_agent_db_vectors.parquet", split="train")
|
|
185
|
+
|
|
186
|
+
# Add native FAISS index for sub-millisecond similarity search
|
|
187
|
+
dataset.add_faiss_index(column="embedding")
|
|
188
|
+
scores, samples = dataset.get_nearest_examples("embedding", query_vector, k=5)
|
|
189
|
+
```
|
|
190
|
+
|
|
191
|
+
### 2. 🗄️ Full Universal SQL Database (`open_agent_db_sql.sqlite.zst`)
|
|
192
|
+
The complete 30+ GB offline SQLite database with full-text search (FTS5), schema indices, and raw BLOB files for 3.48M+ assets, compressed down to ~12 GB with Zstandard:
|
|
193
|
+
|
|
194
|
+
```bash
|
|
195
|
+
# Download compressed database via huggingface-cli
|
|
196
|
+
hf download ns0bj/open-agent-db open_agent_db_sql.sqlite.zst --repo-type dataset --local-dir ./data
|
|
197
|
+
|
|
198
|
+
# Decompress to pristine 31 GB SQLite database (takes ~10 seconds)
|
|
199
|
+
unzstd ./data/open_agent_db_sql.sqlite.zst -o ./data/open_agent_db.sqlite
|
|
200
|
+
|
|
201
|
+
# Query immediately with SQLite or DuckDB
|
|
202
|
+
sqlite3 ./data/open_agent_db.sqlite "SELECT count(*) FROM assets;"
|
|
203
|
+
```
|
|
204
|
+
|
|
205
|
+
---
|
|
206
|
+
|
|
207
|
+
## 🌐 Web Explorer
|
|
208
|
+
|
|
209
|
+
Open-Agent-DB includes a fast web explorer located in [`web/`](web/):
|
|
210
|
+
|
|
211
|
+
* **Static Mode**: Runs 100% on GitHub Pages without a backend.
|
|
212
|
+
* **Local Mode**: Point it to your local SQLite database for instantaneous queries across all 3.48M assets.
|
|
213
|
+
|
|
214
|
+
To run locally:
|
|
215
|
+
```bash
|
|
216
|
+
cd web && npx serve .
|
|
217
|
+
```
|
|
218
|
+
|
|
219
|
+
---
|
|
220
|
+
|
|
221
|
+
## ❓ FAQ
|
|
222
|
+
|
|
223
|
+
### Where can I find a list of all MCP servers?
|
|
224
|
+
Open-Agent-DB merges 113,000+ Model Context Protocol servers from the official MCP Registry, Glama, Smithery and Awesome MCP lists into one searchable SQLite database, with install commands and Claude Desktop / Cursor config snippets. Browse the top ones on the [MCP server database page](https://nsozturk.github.io/open-agent-db/mcp-servers.html).
|
|
225
|
+
|
|
226
|
+
### How do I find Claude Code skills, subagents and CLAUDE.md examples?
|
|
227
|
+
Search the index of `SKILL.md` skills, `.claude/agents/*.md` subagents and `CLAUDE.md` / `AGENTS.md` project rules; each result links to its source repository and install command.
|
|
228
|
+
|
|
229
|
+
### Is there a dataset of Cursor rules (`.cursorrules` / `.mdc`)?
|
|
230
|
+
Yes. Legacy `.cursorrules` files and `.cursor/rules/*.mdc` trigger rules are indexed with author, stars and full text.
|
|
231
|
+
|
|
232
|
+
### Can I download the whole database?
|
|
233
|
+
Yes — the zstd-compressed SQLite database and the Parquet vector embeddings are free on [Hugging Face](https://huggingface.co/datasets/ns0bj/open-agent-db) under the MIT license.
|
|
234
|
+
|
|
235
|
+
### Does it work offline?
|
|
236
|
+
Yes. After downloading, full-text search (FTS5) and semantic vector search (FAISS-ready 128-d embeddings) run locally with no API keys.
|
|
237
|
+
|
|
238
|
+
---
|
|
239
|
+
|
|
240
|
+
## 🔁 Alternatives & how it compares
|
|
241
|
+
|
|
242
|
+
Open-Agent-DB is an open-source, downloadable alternative to browsing MCP and skill directories one by one. It aggregates the [official MCP Registry](https://github.com/modelcontextprotocol/registry), Glama, Smithery and the Awesome MCP lists, plus agent skills and rules from GitHub, into a single offline dataset. Use those registries for hosted discovery; use Open-Agent-DB when you want everything in one SQL- and vector-searchable file.
|
|
243
|
+
|
|
244
|
+
---
|
|
245
|
+
|
|
246
|
+
## 📑 Citation
|
|
247
|
+
|
|
248
|
+
If you use the dataset, cite it via [`CITATION.cff`](CITATION.cff).
|
|
249
|
+
|
|
250
|
+
---
|
|
251
|
+
|
|
252
|
+
## 📄 License
|
|
253
|
+
|
|
254
|
+
This project is licensed under the [MIT License](LICENSE) - see the LICENSE file for details.
|
|
255
|
+
|
|
256
|
+
Developed and maintained by **Enes Öztürk** ([@nsozturk](https://github.com/nsozturk)).
|
|
@@ -0,0 +1,227 @@
|
|
|
1
|
+
<p align="center">
|
|
2
|
+
<img src="https://raw.githubusercontent.com/nsozturk/open-agent-db/main/assets/banner.png" alt="Open-Agent-DB — The Universal AI Agent & MCP Hub" width="100%">
|
|
3
|
+
</p>
|
|
4
|
+
|
|
5
|
+
<div align="center">
|
|
6
|
+
|
|
7
|
+
<p>
|
|
8
|
+
<a href="https://www.npmjs.com/package/open-agent-db"><img src="https://img.shields.io/npm/v/open-agent-db?style=flat-square&color=cb3837&logo=npm" alt="NPM Version" /></a>
|
|
9
|
+
<img src="https://img.shields.io/badge/python-3.8%2B-blue?style=flat-square&logo=python&logoColor=white" alt="Python" />
|
|
10
|
+
<img src="https://img.shields.io/badge/node-%3E%3D18.0-339933?style=flat-square&logo=nodedotjs&logoColor=white" alt="Node" />
|
|
11
|
+
<img src="https://img.shields.io/badge/total%20assets-3.48M%2B-purple?style=flat-square" alt="Total Assets" />
|
|
12
|
+
<img src="https://img.shields.io/badge/mcp%20servers-113K%2B-orange?style=flat-square" alt="MCP Servers" />
|
|
13
|
+
<img src="https://img.shields.io/badge/vector%20db-Parquet%20Dense%20Embeddings-success?style=flat-square" alt="Vector DB" />
|
|
14
|
+
<img src="https://img.shields.io/badge/database-SQLite%20FTS5-blue?style=flat-square&logo=sqlite&logoColor=white" alt="Database" />
|
|
15
|
+
<a href="https://huggingface.co/datasets/ns0bj/open-agent-db"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Datasets%20%26%20Vectors-yellow?style=flat-square" alt="Hugging Face" /></a>
|
|
16
|
+
<a href="https://evren.ssyz.org.tr/datasets/u4be40c9fa813/open-agent-db"><img src="https://img.shields.io/badge/EVREN-NLP%20Dataset-14b8a6?style=flat-square" alt="EVREN dataset" /></a>
|
|
17
|
+
<a href="AGENTS.md"><img src="https://img.shields.io/badge/AI%20Agent-Native%20Ready-blueviolet?style=flat-square" alt="AI Agent Ready" /></a>
|
|
18
|
+
<a href="https://nsozturk.github.io/open-agent-db/"><img src="https://img.shields.io/badge/docs-GitHub%20Pages-00bcd4?style=flat-square" alt="Docs" /></a>
|
|
19
|
+
<img src="https://img.shields.io/badge/license-MIT-green?style=flat-square" alt="License" />
|
|
20
|
+
</p>
|
|
21
|
+
|
|
22
|
+
<p align="center">
|
|
23
|
+
<code>ai-agents</code> •
|
|
24
|
+
<code>model-context-protocol</code> •
|
|
25
|
+
<code>mcp-servers</code> •
|
|
26
|
+
<code>claude-skills</code> •
|
|
27
|
+
<code>cursor-rules</code> •
|
|
28
|
+
<code>vector-database</code> •
|
|
29
|
+
<code>semantic-search</code> •
|
|
30
|
+
<code>sqlite-fts5</code> •
|
|
31
|
+
<code>faiss</code>
|
|
32
|
+
</p>
|
|
33
|
+
|
|
34
|
+
**The Universal Open Database and Semantic Vector Search Engine for AI Agent Skills, MCP Servers, Project Directives, and Rules.**
|
|
35
|
+
|
|
36
|
+
[Documentation](https://nsozturk.github.io/open-agent-db/) • [Features](#-features) • [Ecosystems](#-supported-ecosystems) • [Quick Start](#-quick-start) • [Vector Search](#-semantic-vector-search) • [Hugging Face](#-hugging-face-integration) • [License](#-license)
|
|
37
|
+
|
|
38
|
+
</div>
|
|
39
|
+
|
|
40
|
+
---
|
|
41
|
+
|
|
42
|
+
# Open-Agent-DB — Open Dataset & Search Engine for AI Agent Skills and MCP Servers
|
|
43
|
+
|
|
44
|
+
**Open-Agent-DB is a free, open-source dataset and offline search engine for AI agent skills, MCP servers and coding-agent rules: 3.48M+ assets from Claude Code, Cursor, Codex, Gemini, Copilot and OpenCode in SQLite FTS5 and Parquet vectors.**
|
|
45
|
+
|
|
46
|
+
📚 [Docs](https://nsozturk.github.io/open-agent-db/) · 🔌 [MCP server database](https://nsozturk.github.io/open-agent-db/mcp-servers.html) · 🧠 [Claude Code skills](https://nsozturk.github.io/open-agent-db/claude-code-skills.html) · 🎯 [Cursor rules](https://nsozturk.github.io/open-agent-db/cursor-rules.html) · 🤗 [Dataset on Hugging Face](https://huggingface.co/datasets/ns0bj/open-agent-db)
|
|
47
|
+
|
|
48
|
+
## 🌟 What is Open-Agent-DB?
|
|
49
|
+
|
|
50
|
+
**Open-Agent-DB** is the comprehensive, offline-first open catalog and search engine for the AI Agent era. It unifies **3,480,000+ Agent Assets** and **113,000+ MCP (Model Context Protocol) Servers & Tools** across every major ecosystem:
|
|
51
|
+
|
|
52
|
+
* **Universal Agent Skills**: Canonical `SKILL.md` definitions, multi-step actions, and tool integration scripts.
|
|
53
|
+
* **Project System Rules**: `CLAUDE.md`, `AGENTS.md`, `GEMINI.md`, and `copilot-instructions.md`.
|
|
54
|
+
* **Agent Briefs & Personas**: Anthropic Claude Code (`.claude/agents/*.md`), OpenAI Codex (`.codex/agents/*.md`).
|
|
55
|
+
* **Context & Trigger Rules**: Cursor IDE rules (`.cursorrules`, `.cursor/rules/*.mdc`).
|
|
56
|
+
* **MCP Workflows & Servers**: Official Anthropic MCP Registry, Glama, Smithery, OpenCode (`opencode.json`, `mcp.json`).
|
|
57
|
+
* **Curated System Prompts**: High-impact operational templates and task instructions.
|
|
58
|
+
|
|
59
|
+
Every asset is indexed with author metadata, star ratings, and install commands, and stored with full source code inside SQLite FTS5 databases and dense **vector embeddings**.
|
|
60
|
+
|
|
61
|
+
---
|
|
62
|
+
|
|
63
|
+
## ⚡ Features
|
|
64
|
+
|
|
65
|
+
* 🔍 **Sub-second Full-Text Search (FTS5)**: Query 3.48M+ assets instantly via CLI or web.
|
|
66
|
+
* 🧠 **Semantic Vector Search**: Natural language similarity search using dense vector embeddings over skills, tools, and directives.
|
|
67
|
+
* ⚡ **Zero-Install `npx` Run**: Try instantly without downloading or configuring anything.
|
|
68
|
+
* 🚀 **1-Click Agent Installer**:
|
|
69
|
+
* Automatically installs skills directly to `~/.claude/skills/`, `~/.codex/skills/`, or `.cursor/rules/`.
|
|
70
|
+
* Generates ready-to-use JSON config for Claude Desktop (`claude_desktop_config.json`) and Cursor.
|
|
71
|
+
* 🌐 **Dual-Mode Web Explorer**:
|
|
72
|
+
* **GitHub Pages Static Mode**: Zero backend required! Runs 100% in the browser using a client-side curated index.
|
|
73
|
+
* **Local Mode**: Runs against the full SQLite database for unlimited deep searching.
|
|
74
|
+
* 📦 **Hugging Face Hub Native**: One-command pipeline to export metadata or dense vector embeddings directly into Hugging Face Datasets (`Parquet` / `Arrow`).
|
|
75
|
+
|
|
76
|
+
---
|
|
77
|
+
|
|
78
|
+
## 🧩 Supported Ecosystems
|
|
79
|
+
|
|
80
|
+
| Ecosystem | Primary Target Files | Asset Kinds | Description |
|
|
81
|
+
| :--- | :--- | :--- | :--- |
|
|
82
|
+
| **Universal** | `SKILL.md`, `AGENTS.md`, `AGENT.md` | `skill`, `project_rules` | Cross-platform agent rules and tool skills |
|
|
83
|
+
| **Claude Code** | `CLAUDE.md`, `.claude/agents/*.md`, `.claude/skills/*` | `project_rules`, `agent_brief`, `skill` | Anthropic Claude Code CLI configurations & personas |
|
|
84
|
+
| **Cursor** | `.cursorrules`, `.cursor/rules/*.mdc` | `project_rules`, `trigger_rule` | Cursor IDE rules and trigger-based `.mdc` specs |
|
|
85
|
+
| **OpenAI Codex** | `.codex/agents/*.md`, `.codex-plugin/*`, `mcp.json` | `agent_brief`, `workflow`, `project_rules` | OpenAI Codex CLI agents and plugin specifications |
|
|
86
|
+
| **Google Gemini** | `GEMINI.md`, `.gemini/config/skills/*` | `project_rules`, `skill` | Google Antigravity & Gemini CLI directives |
|
|
87
|
+
| **GitHub Copilot**| `copilot-instructions.md`, `.github/copilot-*` | `project_rules` | GitHub Copilot Workspace custom guidelines |
|
|
88
|
+
| **OpenCode** | `opencode.json`, `.opencode/*` | `workflow`, `skill` | OpenCode agent workflows and tool configs |
|
|
89
|
+
|
|
90
|
+
---
|
|
91
|
+
|
|
92
|
+
## 🚀 Quick Start
|
|
93
|
+
|
|
94
|
+
### 1. Instant Run (Zero Installation via `npx`)
|
|
95
|
+
|
|
96
|
+
You can run Open-Agent-DB immediately using Node.js without installing anything:
|
|
97
|
+
|
|
98
|
+
```bash
|
|
99
|
+
# Search across all indexed assets
|
|
100
|
+
npx open-agent-db search "postgres"
|
|
101
|
+
|
|
102
|
+
# Inspect MCP server details & install command
|
|
103
|
+
npx open-agent-db info "Prisma Postgres"
|
|
104
|
+
|
|
105
|
+
# Generate Claude Desktop MCP JSON config snippet
|
|
106
|
+
npx open-agent-db install "Prisma Postgres"
|
|
107
|
+
```
|
|
108
|
+
|
|
109
|
+
### 2. Global Installation (NPM or Pip)
|
|
110
|
+
|
|
111
|
+
#### Via NPM (Node.js ecosystem):
|
|
112
|
+
```bash
|
|
113
|
+
npm install -g open-agent-db
|
|
114
|
+
open-agent search "docker"
|
|
115
|
+
```
|
|
116
|
+
|
|
117
|
+
#### Via Pip (Python & AI ecosystem):
|
|
118
|
+
```bash
|
|
119
|
+
pip install open-agent-db
|
|
120
|
+
open-agent search "react"
|
|
121
|
+
```
|
|
122
|
+
|
|
123
|
+
---
|
|
124
|
+
|
|
125
|
+
## 🧠 Semantic Vector Search
|
|
126
|
+
|
|
127
|
+
Open-Agent-DB includes a native Vector Search Engine that allows natural language semantic queries:
|
|
128
|
+
|
|
129
|
+
```bash
|
|
130
|
+
# Natural language semantic query
|
|
131
|
+
python3 vector_search.py "automate database schema migrations with safety checks"
|
|
132
|
+
|
|
133
|
+
# Search for specific tools with top-K limit
|
|
134
|
+
python3 vector_search.py "ios swift metal shader visual effects" --top-k 5
|
|
135
|
+
|
|
136
|
+
# Rebuild local vector embeddings index from SQLite
|
|
137
|
+
python3 vector_search.py --build --limit 50000
|
|
138
|
+
```
|
|
139
|
+
|
|
140
|
+
---
|
|
141
|
+
|
|
142
|
+
## 🤗 Hugging Face Integration
|
|
143
|
+
|
|
144
|
+
Open-Agent-DB provides a unified Hub on Hugging Face Datasets ([`ns0bj/open-agent-db`](https://huggingface.co/datasets/ns0bj/open-agent-db)) hosting two core databases:
|
|
145
|
+
|
|
146
|
+
The metadata catalogue (3.65M unified rows + 150K MCP registry items, metadata only, secrets redacted) is also published on Türkiye's EVREN AI platform as an NLP dataset: [`open-agent-db` on EVREN](https://evren.ssyz.org.tr/datasets/u4be40c9fa813/open-agent-db) — uploaded with [evren-platform-api](https://github.com/nsozturk/evren-platform-api).
|
|
147
|
+
|
|
148
|
+
### 1. 🧠 Semantic Vector DB (`open_agent_db_vectors.parquet`)
|
|
149
|
+
Pre-computed dense 128-dimensional float32 vector embeddings over Agent Skills, MCP Servers, and Directives. Ready for instant semantic similarity search via FAISS, cosine distance, or LanceDB.
|
|
150
|
+
|
|
151
|
+
```python
|
|
152
|
+
from datasets import load_dataset
|
|
153
|
+
|
|
154
|
+
# Load vector embeddings directly from Hugging Face
|
|
155
|
+
dataset = load_dataset("ns0bj/open-agent-db", data_files="open_agent_db_vectors.parquet", split="train")
|
|
156
|
+
|
|
157
|
+
# Add native FAISS index for sub-millisecond similarity search
|
|
158
|
+
dataset.add_faiss_index(column="embedding")
|
|
159
|
+
scores, samples = dataset.get_nearest_examples("embedding", query_vector, k=5)
|
|
160
|
+
```
|
|
161
|
+
|
|
162
|
+
### 2. 🗄️ Full Universal SQL Database (`open_agent_db_sql.sqlite.zst`)
|
|
163
|
+
The complete 30+ GB offline SQLite database with full-text search (FTS5), schema indices, and raw BLOB files for 3.48M+ assets, compressed down to ~12 GB with Zstandard:
|
|
164
|
+
|
|
165
|
+
```bash
|
|
166
|
+
# Download compressed database via huggingface-cli
|
|
167
|
+
hf download ns0bj/open-agent-db open_agent_db_sql.sqlite.zst --repo-type dataset --local-dir ./data
|
|
168
|
+
|
|
169
|
+
# Decompress to pristine 31 GB SQLite database (takes ~10 seconds)
|
|
170
|
+
unzstd ./data/open_agent_db_sql.sqlite.zst -o ./data/open_agent_db.sqlite
|
|
171
|
+
|
|
172
|
+
# Query immediately with SQLite or DuckDB
|
|
173
|
+
sqlite3 ./data/open_agent_db.sqlite "SELECT count(*) FROM assets;"
|
|
174
|
+
```
|
|
175
|
+
|
|
176
|
+
---
|
|
177
|
+
|
|
178
|
+
## 🌐 Web Explorer
|
|
179
|
+
|
|
180
|
+
Open-Agent-DB includes a fast web explorer located in [`web/`](web/):
|
|
181
|
+
|
|
182
|
+
* **Static Mode**: Runs 100% on GitHub Pages without a backend.
|
|
183
|
+
* **Local Mode**: Point it to your local SQLite database for instantaneous queries across all 3.48M assets.
|
|
184
|
+
|
|
185
|
+
To run locally:
|
|
186
|
+
```bash
|
|
187
|
+
cd web && npx serve .
|
|
188
|
+
```
|
|
189
|
+
|
|
190
|
+
---
|
|
191
|
+
|
|
192
|
+
## ❓ FAQ
|
|
193
|
+
|
|
194
|
+
### Where can I find a list of all MCP servers?
|
|
195
|
+
Open-Agent-DB merges 113,000+ Model Context Protocol servers from the official MCP Registry, Glama, Smithery and Awesome MCP lists into one searchable SQLite database, with install commands and Claude Desktop / Cursor config snippets. Browse the top ones on the [MCP server database page](https://nsozturk.github.io/open-agent-db/mcp-servers.html).
|
|
196
|
+
|
|
197
|
+
### How do I find Claude Code skills, subagents and CLAUDE.md examples?
|
|
198
|
+
Search the index of `SKILL.md` skills, `.claude/agents/*.md` subagents and `CLAUDE.md` / `AGENTS.md` project rules; each result links to its source repository and install command.
|
|
199
|
+
|
|
200
|
+
### Is there a dataset of Cursor rules (`.cursorrules` / `.mdc`)?
|
|
201
|
+
Yes. Legacy `.cursorrules` files and `.cursor/rules/*.mdc` trigger rules are indexed with author, stars and full text.
|
|
202
|
+
|
|
203
|
+
### Can I download the whole database?
|
|
204
|
+
Yes — the zstd-compressed SQLite database and the Parquet vector embeddings are free on [Hugging Face](https://huggingface.co/datasets/ns0bj/open-agent-db) under the MIT license.
|
|
205
|
+
|
|
206
|
+
### Does it work offline?
|
|
207
|
+
Yes. After downloading, full-text search (FTS5) and semantic vector search (FAISS-ready 128-d embeddings) run locally with no API keys.
|
|
208
|
+
|
|
209
|
+
---
|
|
210
|
+
|
|
211
|
+
## 🔁 Alternatives & how it compares
|
|
212
|
+
|
|
213
|
+
Open-Agent-DB is an open-source, downloadable alternative to browsing MCP and skill directories one by one. It aggregates the [official MCP Registry](https://github.com/modelcontextprotocol/registry), Glama, Smithery and the Awesome MCP lists, plus agent skills and rules from GitHub, into a single offline dataset. Use those registries for hosted discovery; use Open-Agent-DB when you want everything in one SQL- and vector-searchable file.
|
|
214
|
+
|
|
215
|
+
---
|
|
216
|
+
|
|
217
|
+
## 📑 Citation
|
|
218
|
+
|
|
219
|
+
If you use the dataset, cite it via [`CITATION.cff`](CITATION.cff).
|
|
220
|
+
|
|
221
|
+
---
|
|
222
|
+
|
|
223
|
+
## 📄 License
|
|
224
|
+
|
|
225
|
+
This project is licensed under the [MIT License](LICENSE) - see the LICENSE file for details.
|
|
226
|
+
|
|
227
|
+
Developed and maintained by **Enes Öztürk** ([@nsozturk](https://github.com/nsozturk)).
|