AGNTCY Integration Capabilities: An Overview

by Ramiz Polic, Luca Muscariello

layout: statement

Goals

Support Agentic Standards

Support widely-adopted standards for agentic development like MCP servers

Enhance Developer Experience

Support AI-assisted developer workflows via tools and IDEs like Visual Studio Code

Simplified Integration

Easy integration and usage of AGNTCY components like OASF and ADS

layout: default

Methodology

Schema Extensions
Use OASF records as data carriers for common tools
Add native support for LLMs, Prompts, A2A Cards, MCP servers
Data Extractors and Transformers
Generate tool-specific configuration files from OASF records
Support tool usage via VS Code Copilot and Continue.Dev
Workflow Integration
Provide simple setup instructions
Enable immediate productivity

layout: center

Architecture

flowchart TD
    subgraph EX[Schema Extensions]
        C1[MCP Servers]
        C2[LLM Models]
        C3[A2A Cards]
        C4[Prompts]
    end

    subgraph SRC[OASF Record sources]
        H1[Hub Service]
        H2[ADS Network]
    end

    EX -.-> |*defines extension schema for data attached to a record*| A[OASF Record]

    SRC --> |record found and retreived from a given source| A

    subgraph INT[Tooling Integrations]
      D[Data Extractors & Transformers] --> |generates VSCode configs| E1[VSCode Support]
      D[Data Extractors & Transformers] --> |generates Continue configs| E2[Continue Support]
    end

    A --> |record ingested locally to support| INT

    style A fill:#93c5fd,stroke:#3b82f6
    style D fill:#fbcfe8,stroke:#ec4899
    style EX stroke-dasharray: 5 5

background: "#eff3fc" layout: center

OASF Schema Extensions

layout: two-cols-header

OASF Schema: LLM Extension

::left::

{
  "extensions": [{
    "name": "schema.oasf.agntcy.org/features/runtime/model",
    "version": "v1.0.0",
    "data": {
      "models": [
        {
          "model": "deepseek-r1:1.5b",
          "provider": "ollama",
          "api_base": "http://localhost:11434",
          "prompt": "You are an expert software developer..."
        },
        {
          "model": "gpt-4o",
          "provider": "azure",
          "api_base": "${input:AZURE_OPENAI_API_BASE}",
          "api_key": "${input:AZURE_OPENAI_API_KEY}",
        }
      ]
    }
  }]
}

::right::

## Features Describes LLM support and its configuration for a given agent.
- Multi-model support - Local and cloud providers - Model tuning parameters - Compatible with existing tools

layout: two-cols-header

OASF Schema: LLM Prompt Extensions

::left::

{
  "extensions": [{
    "name": "schema.oasf.agntcy.org/features/runtime/prompt",
    "version": "v1.0.0",
    "data": {
      "prompts": [
        {
          "name": "PR Summary",
          "description": "PR analysis",
          "prompt": "Summarize the pull request in detail..."
        },
        {
          "name": "PR Review",
          "description": "PR review",
          "prompt": "Review the pull request in detail..."
        }
      ]
    }
  }]
}

::right::

## Features Describes common LLM interaction prompts to use the agent.
- Structured LLM Prompts - Categorized by purpose - Descriptive metadata - Task-specific instructions - Compatible with existing tools

layout: two-cols-header

OASF Schema: MCP Server Extension

::left::

{
  "extensions": [{
    "name": "schema.oasf.agntcy.org/features/runtime/mcp",
    "version": "v1.0.0",
    "data": {
      "servers": {
        "github": {
          "command": "docker",
          "args": [
            "run", "-i", "--rm", 
            "-e", "GITHUB_PAT",
            "ghcr.io/github/github-mcp-server"
          ],
          "env": {
            "GITHUB_PAT": "${input:GITHUB_PAT}"
          }
        }
      }
    }
  }]
}

::right::

## Features Describes MCP servers required to run and interact with the agent.
- Support for multiple servers - Input variable mapping and templating - Compatible with existing tools

layout: two-cols-header

OASF Schema: A2A Extensions

::left::

{
  "extensions": [{
    "name": "schema.oasf.agntcy.org/features/runtime/a2a",
    "version": "v1.0.0",
    "data": {
      "name": "example-agent",
      "description": "An agent that performs web searches",
      "url": "http://localhost:8000",
      "capabilities": {
        "streaming": true,
        "pushNotifications": false
      },
      "defaultInputModes": ["text"],
      "defaultOutputModes": ["text"],
      "skills": [
        { "id": "browser", "name": "browser automation" }
      ]
    }
  }]
}

::right::

## Features Describes A2A card details for communication and its usage with A2A protocol.
- Common A2A Card schema - Compatible with existing tools

background: "#eff3fc" layout: center

Agentic Workflow: IDE Integrations

layout: default

Agentic Workflow: Example OASF record

<<< @/integrations/record.json

layout: two-cols-header

VS Code Integration

Explores ways to use OASF records to enable its usage in agentic workflows with VS Code.

::left::

Implementation

Load and Process OASF records
Extract data from OASF record about:
- MCP servers
- NOTE: other components are not yet supported
Generate .vscode/mcp.json configuration file

::right::

## Features - **MCP Server Configuration** - Automatic server setup - Secure credential handling - Input variable templating - **Copilot Agent Mode Support** - Native integration - Interactive credential prompts

background: "#eff3fc" layout: center

VS Code Integration: Live Demo

layout: two-cols-header

Continue Integration

Explores ways to use OASF records to enable its usage in agentic workflows with VS Code Continue.dev extension.

::left::

Implementation

Load and Process OASF records
Extract data from OASF record about:
- MCP servers
- LLM Data
- Prompts
- A2A information
Generate .continue/assistants/
configuration file

::right::

Features

Quick Provisioning
Multi-model configuration
Pre-defined prompts
MCP server integration
Dev Experience
Assistant gallery integration
Model provider selection
Prompt usage

background: "#eff3fc" layout: center

Continue Integration: Live Demo

layout: two-cols-header

MCP to OASF Example: Agentic App

Demonstrate the usage of OASF agents by loading them into IDEs and performing an agentic workflow to scan MCP server records from GitHub and create matching OASF records.

::left::

VSCode Workflow

Open VSCode chat console
Switch to LLM (e.g., Claude)
Enable Agent mode
Enter conversion generation prompt
Check results

::right::

Continue Workflow

Open Continue chat console
Refresh Assistants tab
Select OASF-generated assistant
Switch to configured LLM (Azure GPT-4o)
Select conversion generation prompt
Check results

layout: default

MCP to OASF Example: OASF Record

Connecting everything together to showcase an example application.
Built on top of OASF and ADS with native IDE support.

<<< @/mcp-extractor-agent/extractor.record.json

background: "#eff3fc" layout: center

MCP to OASF Example: Live Demo

layout: statement

Outcomes

Extensible Schema: OASF can easily support third-party integrations via extensions

Minimal Impact: No significant changes needed to ADS or OASF projects

Simple Integration: Straightforward integration of AGNTCY components

Developer Experience: Leverages familiar tools with minimal configuration

Transformation Pattern: Data extraction and transformation approach works well

layout: end

Thank You

github.com/agntcy/dir