Jentic SDK & Management Component Proxy [Experimental]

Jentic furnishes builders of intelligent agents with the capability to quickly ascertain and integrate third-party Application Programming Interfaces and operational sequences—bypassing the necessity to author or maintain any interface-specific source code.

This unified repository houses:

The Jentic Software Development Kit (SDK) – A Python package designed for querying, loading definitions, and invoking remote APIs or workflows, alongside utilities for transforming these actions into constructs usable by Large Language Models (LLMs).
The Jentic Management Component Proxy (MCP) Module – An MCP endpoint that mirrors the functionality of the SDK, accessible by any client adhering to the MCP specification (e.g., Windsurf, Claude Desktop, Cursor, etc.).

Refer to the distinct README documentation for comprehensive specifics:

python/README.md – Detailed SDK operational guidance & function catalog
mcp/README.md – Instructions for establishing and configuring the MCP server instance

The underlying data empowering the SDK is sourced from the Jentic Public APIs Artifacts repository.

Initial Setup

1. Install via Python Package Manager

bash pip install jentic

2. Secure Your Agent Authentication Token

Navigate to https://app.jentic.com/sign-in to provision an agent and retrieve the corresponding access credential.

bash export JENTIC_AGENT_API_KEY=

3. Leverage the SDK

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

async def main(): interface_client = Jentic()

# 1️⃣ Locate a required capability
discovery_results = await interface_client.search(SearchRequest(query="dispatch a message to Discord"))
unique_identifier = discovery_results.results[0].id  # Format is op_... or wf_...

# 2️⃣ Retrieve definition details (examine schemas / authentication requirements, view input parameters for functions)
details_response = await interface_client.load(LoadRequest(ids=[unique_identifier]))
required_inputs = details_response.tool_info[unique_identifier].inputs
print (required_inputs)

# 3️⃣ Execute the capability
operation_result = await interface_client.execute(
    ExecutionRequest(id=unique_identifier, inputs={"destination_identifier": "123", "message_body": "Greetings!"})
)
print(operation_result)

asyncio.run(main())

4. Integration with Cognitive Agent Architectures (Optional)

If you require fully formalized descriptor artifacts compatible with models from Anthropic or OpenAI, employ the runtime utility functions:

python from jentic.lib.agent_runtime import AgentToolManager

manager = AgentToolManager(format="anthropic") tool_definitions = manager.generate_tool_definitions() # Supply these definitions to the Language Model execution_outcome = await manager.execute_tool("discord_dispatch_message", {"destination_identifier": "123", "message_body": "Hello there"}) print(execution_outcome)

Utilizing the Management Component Proxy (MCP) Endpoint

To make these exact functionalities accessible through an MCP architecture, adhere to the guidance provided in mcp/README.md.

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

Subsequently, configure your designated MCP-compliant client software to direct its communications toward the active server instance (refer to the auxiliary README for configuration examples).