EkaMantra & AUM (ॐ): The Open Standard for Agents

In ancient Indian philosophy, AUM (ॐ) represents the primordial seed sound that manifests creation. In the Liate ecosystem, AUM stands for the Agent Unified Manifest — the declarative standard that manifests autonomous agents from simple JSON into living execution.

Just as package.json unified Node.js packages and Dockerfile unified containerization, liate.json is the open standard for sovereign AI agents.

The 5 Pillars of EkaMantra (L · I · A · T · E)

liate.json organizes an autonomous agent into five clean, orthogonal keys:

1. L — LLM Provider & Model

Defines the intelligence core. Agentantra natively supports local Ollama models (Llama 3.3, DeepSeek-R1), Indian sovereign providers (Sarvam AI), and global frontier APIs:

"L": "sarvam/sarvam-105b"

2. I — Integration & Session Memory

Manages stateful persistence, conversation history, and local vector stores. Memory stays strictly isolated to the user's host:

"I": {
  "memory": "sessions/analyst-session-01"
}

3. A — Agent Identity, Intent & Skills

Contains the agent's name, system prompt, goal boundary, and external markdown skill sheets:

"A": {
  "name": "data-analyst",
  "intent": "Perform financial vector analytics on Indian equities.",
  "skills": "./skills.md"
}

4. T — Tools & Universal MCP Servers

Declares external capabilities via Anthropic's Model Context Protocol (MCP) or native binary plugins:

"T": {
  "postgres-tool": {
    "command": "npx",
    "args": ["-y", "@modelcontextprotocol/server-postgres"]
  }
}

5. E — Environment & Security Guardrails

Defines environment variables, runtime secrets, and max execution turn boundaries:

"E": {
  "MAX_TURNS": "15",
  "DB_URI": "$SOVEREIGN_DB_URI"
}

Why Declarative Standardization Wins

• Version Controlled: Agents are committed to Git repositories and reviewed through standard Pull Requests.
• Zero Lock-In: Move agents effortlessly from local developer machines to dedicated Mumbai sovereign cloud clusters without changing a single line of application code.
• Security Auditability: Security teams can inspect an agent's entire permission matrix and tool access in under 10 seconds.