Jentic SDK & MCP Integration Module [Preview]

Jentic provides AI-agent developers with the means to seamlessly incorporate external service endpoints and business logic flows—eliminating the burden of manual API code crafting and upkeep.

This unified repository houses:

The Jentic Core Library (SDK) – A Python package for locating, loading, and invoking remote interfaces/workflows, featuring utilities to adapt these actions into consumable constructs for Large Language Models (LLMs).
The Jentic MCP Connector – A service layer implementing the Machine Communication Protocol (MCP) standard, offering the identical functionality to any client adhering to MCP specifications (e.g., Windsurf, Claude Desktop, Cursor, etc.).

Refer to the specific documentation files for comprehensive details:

python/README.md – Detailed SDK utilization guidelines and reference for all exposed methods.
mcp/README.md – Instructions for deploying and configuring the MCP gateway service.

The system's operational knowledge base is sourced from the catalogue within the Jentic Public API Registry.

Expedited Setup

1. Install Python Dependency

bash pip install jentic

2. Secure Your Agent Credential

Obtain your authorization token by registering at https://app.jentic.com/sign-in and copying the generated key.

bash export JENTIC_AGENT_API_KEY=

3. Employing the Core Library

python import asyncio from jentic import Jentic, SearchRequest, LoadRequest, ExecutionRequest

async def initialize_and_run(): interface_client = Jentic()

# 1️⃣ Identify a needed function
search_outcomes = await interface_client.search(SearchRequest(query="dispatch a message via Discord"))
operation_identifier = search_outcomes.results[0].id  # Will be formatted as op_... or wf_...

# 2️⃣ Retrieve metadata (examine required parameters, authentication methods, input schemas)
metadata_response = await interface_client.load(LoadRequest(ids=[operation_identifier]))
required_parameters = metadata_response.tool_info[operation_identifier].inputs
print (required_parameters)

# 3️⃣ Execute the operation
execution_result = await interface_client.execute(
    ExecutionRequest(id=operation_identifier, inputs={"recipient_id": "123", "message_body": "Greetings!"})
)
print(execution_result)

asyncio.run(initialize_and_run())

4. Integration with Cognitive Agents (Optional)

For generating fully structured descriptor files compatible with models from Anthropic or OpenAI, utilize the included runtime utilities:

python from jentic.lib.agent_runtime import AgentToolManager

manager = AgentToolManager(format="anthropic") tool_definitions = manager.generate_tool_definitions() # Supply these definitions to your LLM instance execution_feedback = await manager.execute_tool("discord_send_message", {"recipient_id": "123", "content": "Salutations"}) print(execution_feedback)

Utilizing the MCP Gateway Module

To expose these functionalities over the MCP protocol, follow the setup guide found in mcp/README.md.

bash uvx --from \ git+https://github.com/jentic/jentic-sdks.git@main#subdirectory=mcp \ mcp

Subsequently, configure your MCP-compliant application layer to direct its communication endpoint towards the actively running server instance (check the subdirectory README for example client configuration snippets).