Peta Documentation
The Control Plane for MCP
Technical Overview
One layer between your AI agents and your tools & APIs — policies enforced, credentials injected server-side, every call logged.

The Peta platform consists of three integrated components that work together to provide complete MCP governance:
Peta Core
Managed MCP Runtime & Zero-Trust Gateway
The heart of Peta — manages MCP server lifecycle, injects credentials server-side, and enforces policies at the gateway level.
Peta Console
Policy Engine & Access Control
Web-based administration interface for defining which users and agents can access which tools. Configure RBAC/ABAC policies and monitor usage.
Peta Desk
Human-in-the-Loop Approval System
Desktop app for real-time approval of high-risk operations. Review tool calls before execution with one-click approve/deny.
Reference endpoints for Peta and related repositories:
- https://peta.io – product site
- https://docs.peta.io – documentation home
- https://github.com/dunialabs/peta-core – Peta Core repository
- https://github.com/dunialabs/peta-desk – Peta Desk repository
- https://dunialabs.io – Dunia Labs
Peta MCP Platform supports all standard MCP transport modes, ensuring compatibility with any MCP-enabled application:
StdIO
Standard input/output communication for local process integration and CLI tools
HTTP+SSE
HTTP with Server-Sent Events for real-time bi-directional communication over web protocols
Streamable HTTP
HTTP-based streaming for efficient large data transfer and progressive response handling
Peta implements a comprehensive three-layer security model that ensures zero credential exposure and complete operational control.
Connection Layer
Authentication & Session Management
- • AI agents authenticate using Peta service tokens
- • JWT signature and expiry validation
- • Secure MCP sessions with operation tracking
Processing Layer
Request Validation & Execution
- • RBAC/ABAC permission checks
- • Just-in-time credential decryption (30s TTL)
- • Tool execution with real credentials
- • Sanitized response return
Control Layer
Policy Enforcement Mechanisms
- • Multi-dimensional controls (operation type, data sensitivity)
- • Rule-engine decisions
- • Immutable audit trails
Protocol Compliance
MCP Protocol: 2025-11-25 specification (full compliance)
Encryption Standards
- • AES-256-GCM at rest
- • PBKDF2 (100,000+ iterations)
- • Memory-only decryption with immediate cleanup
- • Credential decryption TTL: 30 seconds
Enterprise Features
- • SSO/SAML support for identity integration
- • Tamper-evident, SIEM-ready logs
- • SOC2 Type II ready, HIPAA capable, GDPR compliant
Platform Support
- • Kubernetes 1.21+
- • Docker containers
- • AWS / GCP / Azure cloud platforms
- • Air-gapped deployment support
Transport Modes
StdIO, HTTP+SSE, Streamable HTTP