# SuperAgentX

> An open-source, modular agentic AI framework that enables AI agents to plan, act, and execute real-world workflows with built-in human approval, governance, and auditability.

SuperAgentX is an open-source Python framework released under the MIT License, built by the superagentxai organization on GitHub. It targets enterprise teams that need to move from prototype to production with autonomous AI agents while maintaining governance, auditability, and human oversight. The project was created in October 2024 and reached its latest release (v1.0.8.3) in May 2026.

## What It Is

SuperAgentX is a multi-agent orchestration framework designed for action-oriented AI workflows rather than conversational chatbots. It provides a unified control plane for tools, models, data, and policies, letting developers compose agents that plan, call tools, interact with browsers and APIs, and execute multi-step workflows—all with persistent state and audit logs. The framework is installable via `pip install superagentx` and requires Python 3.12+.

## Core Capabilities

- **100+ LLMs supported**: OpenAI, Azure OpenAI, Gemini, Claude, Amazon Bedrock, DeepSeek, Ollama, and other open-source models
- **10,000+ MCP (Model Context Protocol) tools**: broad tool ecosystem compatibility
- **Browser Agents**: real browser automation via Playwright for RPA-style workflows
- **Parallel, sequential, or hybrid agent pipelines**: flexible multi-agent communication patterns
- **Contextual memory**: SQL (SQLite, PostgreSQL) and vector database storage for workflow state, agent decisions, tool outputs, and audit logs
- **WebSocket, RESTful API, and IO console interfaces**: multiple deployment surfaces

## Human-in-the-Loop Governance

A built-in Human Approval Governance Agent is a first-class feature of SuperAgentX. When an agent reaches a sensitive action, execution pauses, requests explicit human approval, and then resumes or aborts based on the response. All decisions are persisted for audit. The framework's design principle is that AI cannot act blindly—every autonomous action can be gated by policy-driven access control and traceable approval records.

## Agentic Application Development Lifecycle (ADLC)

The project homepage describes a four-stage ADLC model:
1. **Conversational Application Generation** – submit specs (text, PDFs, Figma designs) to generate agent-powered apps
2. **Agentic Workflow Integration** – embed multi-agent workflows into backends and connect data sources
3. **Policy, Governance & Approvals** – embed IAM policies and human-approval gates
4. **Enterprise Grade Privacy** – PII redaction, data residency controls, and compliance-ready observability

## Out-of-the-Box Specialized Agents

SuperAgentX ships with pre-built agent types: a RAG Agent for querying unstructured documents, an Extract Agent for pulling structured data from raw files, and a Browser RPA Agent for interacting with legacy systems or third-party portals without modern APIs. The framework also supports cognitive LLM Agents for reasoning and deterministic Task Agents for precise execution.

## Update: v1.0.8.3

The latest release is v1.0.8.3, published on May 1, 2026. The repository was last updated on May 20, 2026, indicating active maintenance. The project has accumulated 197 stars and 43 forks since its October 2024 creation. GitHub topics on the repository include agentic-ai, agentic-framework, aigovernance, aisecurity, autonomous-agents, rag-pipeline, memory, and orchestration, reflecting the framework's broad scope.

## Features
- Multi-agent orchestration (parallel, sequential, hybrid)
- 100+ LLM support (OpenAI, Azure OpenAI, Gemini, Claude, Bedrock, Ollama, DeepSeek)
- 10,000+ MCP tool support
- Browser automation via Playwright
- Human-in-the-loop approval governance
- Persistent workflow state and audit logs
- SQLite and PostgreSQL data store support
- Vector database memory for contextual retrieval
- RAG Agent for document querying
- Extract Agent for structured data extraction
- Browser RPA Agent for legacy system automation
- WebSocket, RESTful API, and IO console interfaces
- Policy-driven access control
- PII redaction and data residency controls
- Goal-oriented agents with retry mechanisms

## Integrations
OpenAI, Azure OpenAI, Google Gemini, Anthropic Claude, Amazon Bedrock, DeepSeek, Ollama, Playwright, PostgreSQL, SQLite, MCP (Model Context Protocol)

## Platforms
WINDOWS, LINUX, API, DEVELOPER_SDK, CLI

## Pricing
Open Source

## Version
v1.0.8.3

## Links
- Website: https://superagentx.ai
- Documentation: https://docs.superagentx.ai/introduction
- Repository: https://github.com/superagentxai/superagentx
- EveryDev.ai: https://www.everydev.ai/tools/superagentx