Quickstart Guide

The platform is built around a closed-loop framework. Every goal follows a three-step cycle:

1. 1

   Sign in to your Workspace

   Viksa AI is designed for secure, enterprise environments. Obtain your organization login credentials or access the platform directly using corporate SSO/SAML integrations.

   • Log in via your dedicated URL and create a new project.
   • Retrieve your project access keys from the secure dashboard dashboard.
2. 2

   Build your first Python Agent

   Define your agent in main.py with async functions and the @mcp_endpoint decorator (from the injected ViksaAI module in the UI, or viksa_ai.runtime when developing locally). Each endpoint takes a single payload: Dict[str, Any] and is invoked by the Think→Act→Observe executor at runtime.

   from typing import Any, Dict
   
   # In the dashboard editor use: from .ViksaAI import mcp_endpoint
   from viksa_ai.runtime import mcp_endpoint
   
   @mcp_endpoint(description="Greet someone by name from the payload")
   async def greet(payload: Dict[str, Any]) -> Dict[str, Any]:
       name = payload.get("name") or "there"
       return {"message": f"Hello, {name}! Welcome to Viksa AI."}
   

   List pip dependencies in requirements.txt:

   httpx>=0.27.0
   # Optional for local validation: pip install viksa-ai[dev]
   

   Optionally install viksa-ai and run viksa-agent-validate ./my-agent/ before upload. See the Python SDK guide and Creating agents.

3. 3

   Add a Specialized Agent (Optional)

   The Viksa AI executor excels at chaining multiple agents together to fulfill a single, high-level goal. Add a second server health checker:

   from typing import Any, Dict
   from viksa_ai.runtime import mcp_endpoint
   
   @mcp_endpoint(description="Return a simple health status object")
   async def check_health(payload: Dict[str, Any]) -> Dict[str, Any]:
       return {"status": "ok", "service": payload.get("service", "agent")}
   

4. 4

   Trigger dynamic execution

   Open the **Interactive Chat** in your dashboard, describe the objective in plain English, and watch the executor dynamically plan, coordinate, and execute the steps:

   User: "Greet John and check if our servers are healthy."
   
   Think → Act → Observe:
     1. hello_agent.greet(name="John")
     2. server_health.check_health()

5. 5

   Invoke Webhook Triggers

   Automate goal execution from alerts or third-party webhooks. Configure an active **Trigger**, and POST a JSON payload specifying the objective:

   {
     "event": "incident.opened",
     "severity": "high",
     "service": "payments-api",
     "goal": "Investigate elevated error rate and restart unhealthy pods if needed"
   }

Now that you have deployed your first agentic loop, explore advanced configuration topics: