Try XGEN free for 15 days — nothing to install, runs in your browserStart free trial
PlateerAI Labs

Open Source · Library Gallery

Library Gallery

The open-source libraries behind XGEN. Install them with pip, or try every tool right here in your browser.

ingestionEdit2DocsTry this demo
Edit2Docsingestion
PythonMIT
$pip install xgen_edit2docs

An AI-agent-native document engine that produces natively editable DOCX, XLSX, and PPTX files from a one-line intent, and chat-edits existing ones — real paragraphs, cells, and charts.

ingestionrun in the browser
ingestion
Edit2Docs

/ tools

16 libraries.
One install away.

Ingestionlive demo

Contextifier

Extract and chunk 80+ document formats. Tables, code blocks, and structure preserved for retrieval.

Python
Open demo
Ingestionlive demo

Doc2Chunk

Split documents into context-aware chunks with configurable size and overlap.

Python
Open demo
Ingestionlive demo

f2a

Point at any file, get full statistics and an interactive HTML report. 24+ formats.

Python
Open demo
Knowledgelive demo

Synaptic Memory

Auto-ontology, Hebbian learning, four-stage memory consolidation for long-running agents.

Python
Open demo
Agentlive demo

Googer

Web, images, news, and videos. Typed responses, no scraping gymnastics.

Python
Open demo
Ingestionlive demo

Document Adapter

A unified editing adapter (with MCP server) that lets an LLM open, modify, and write Office and HWPX documents in place.

Python
Open demo
Knowledgelive demo

Knowtology

Build navigable knowledge maps for LLM agents — four-tool exploration, snippet retrieval, and structured recall.

Python
Open demo
Knowledgelive demo

OmniFuse

Backend-agnostic GraphRAG that fuses vector and graph retrieval in a single pass. Zero core deps, pluggable stores.

Python
Open demo
Agentlive demo

Toolint

Lint MCP-compatible, zero-dependency Python agent tools — as a library, a CLI, or an MCP server.

Python
Open demo
Agentlive demo

XGEN Harness

The XGEN harness engine — a config-driven, 10-stage pipeline that owns validation, retry, and termination around any model.

Python
Open demo
Agentlive demo

playwLeft

A browser automation library built for agents — fast, scriptable control of real browser sessions.

Python
Open demo
Utilitylive demo

XGen SDK

A curated Python toolkit of reusable backend building blocks — DB pooling, config, storage, ABAC auth, quota, and a LangChain-free agent harness — consolidated into one canonical package.

Python
Open demo
Knowledgelive demo

XGen Agent Memory

A lightweight memory engine fusing semantic search, BM25 keyword search, and typed-edge graph traversal with an online-learned ranker — all in a single SQLite file.

Python
Open demo
Agentlive demo

XGen Agent Runtime

A manifest-driven agent pipeline with a dual-abstraction architecture (stage x strategy slots), five LLM providers, and native MCP support — every step observable and swappable.

Python
Open demo
Agentlive demo

AN-Web

Executes the web as an actionable state machine instead of rendering pixels for human eyes — every page becomes a structured semantic graph agents can reason over and act upon.

Python
Open demo
Ingestionlive demo

Edit2Docs

An AI-agent-native document engine that produces natively editable DOCX, XLSX, and PPTX files from a one-line intent, and chat-edits existing ones — real paragraphs, cells, and charts.

Python
Open demo

/ Runtime

Runtime

The XGEN engine exports code and runs anywhere. Not a server locked to a platform, but a standalone MCP server you can modify, run, and connect on top of standard ecosystems (PyPI, npm).

IndependenceOpen Architecture

MCP Apps — not a wrapper, a compiler

Most platforms keep the engine running inside and wrap only its surface in MCP. XGEN uses the SDK to compile the workflow and the security policy themselves into standard process code, exported as a standalone MCP server package — build once, run anywhere.

The engine stays inside the platform

Wrapper approachA structure that wraps the MCP surface

MCP WRAPPER

Platform Engine (always-on)

Workflow JSON

Stored inside the platform DB

Output

Platform-bound MCP server

Node · Python

The engine runs always-on inside, with MCP wrapped around the outside

Execution logic becomes an independent asset

Compiler approachCompiled into standard process code

XGEN SDK

Workflow

compiled in

Tool

compiled in

Output

Standalone MCP server package

Node · Python

The SDK compiles the workflow and the security policy themselves into standard process code and exports them

IndependenceRuntime SDK

Runtime SDK — the whole engine behind one SDK API

Compose RAG, workflow, session, and LLMOps in standard Python or Node code. Workflows and policies become assets, packaged into something you can deploy.

1

XGEN engine

RAG, workflow, and LLMOps behind a single SDK API

  • RAGKnowledge retrieval
  • WorkflowWorkflow engine
  • SessionSession management
  • LLMOpsOperations and monitoring
2

SDK export

Compose the two SDKs in standard Python or Node code

  • Standard codeSDK composition
  • Workflow and policyTurned into assets
  • Deployment packageRunnable package built
3

Standalone MCP server

Deploy on its own into any environment you want

  • Container
  • Serverless
  • Edge
4

MCP client

Connect from anywhere, to any client

  • ClaudeMCP Client
  • CursorMCP Client
  • Internal AIIn-house systems
  • Custom AppYour own app
  • Any ClientMCP compatible
ConnectivityRuntime API

Runtime API — deploy anywhere, connect to any client

Deploy standalone into containers, serverless, or the edge — whatever environment you want. The standard MCP interface connects to any client.

ClaudeCursorInternal systemsCustom appsAny MCP-compatible client
Library Gallery · Plateer AI Labs