agentify-ai 0.0.1

No project description provided

✨ 🤖 Agentify AI

Agentify AI is a Python and TypeScript library to develop and deploy an Agent Driven Architecture (ADA).

Agentify AI embraces A2A as MCP (see Ecosystem for alternative solutions) and can be deployed on your laptop or on Kubernetes.

The main features added to A2A and MCP are:

• Agent identity with Delegation
• Long running process with async tools
• Agent authorizations with permissions
• Controls with Guardrails
• Rich and custom user interactions with AG-UI
• Discovery with a Registry
• Observability with OpenTelemetry

An agent has:

• A public Identity.
• Some Guardrails defined by its Manager.
• Some Tools.
• Some Skills gained using the Tools.

An Agent:

• Interacts with other Agents.
• Reports to a single Manager.

An agent collaborates with a another Agent based its Skills.

Identity

Outside Agents Identity is provided with:

• PKCI
• OAuth with automatic registration: Dynamic Client Registration (RFC 7591 / RFC 7592) allows clients (internal services, for example) to automatically register with an authorization server (AuthZ server) via an API, obtain their own client_id (and possibly secret or key pair) programmatically, optionally update or delete their registration later.

Inside Agents Identity is provided with:

• SPIFFE – Secure Production Identity Framework for Everyone - A set of standards for securely identifying software systems in dynamic and heterogeneous environments.
• SPIRE – SPIFFE Runtime Environment - An implementation of SPIFFE that issues and manages identities (SVIDs) for workloads.

Integrations

• VS Code Copilot: Via the A2A edition

Ecosystem

• AG2: https://ag2.ai
• AgentScope: https://agentscope.io
• AGNTCY: https://agntcy.org by Cisco
• Bee AI: https://beeai.dev by IBM
• Fast A2A: https://github.com/pydantic/fasta2a by Pydantic AI.
• Kagent-dev: https://github.com/kagent-dev by Solo, to deploy single Agents or MCP servers on Kubernetes.
• MassGen: https://massgen.ai
• MCP Agent: https://github.com/lastmile-ai/mcp-agent
• OpenAgents: https://openagents.org
• OpenAI Swarm: https://github.com/openai/swarm by OpenAI.