Availablecustom integrationAgent Platforms

Custom Agent Runtime

For organizations with highly specialized requirements, MUJPI provides the foundational APIs and inference capabilities necessary to host custom agent runtimes. Bring your own orchestration logic, memory management, and tool integration frameworks while relying on MUJPI for secure, private intelligence.

Architecture & Data Flow

Your custom runtime operates independently, communicating with MUJPI strictly via authenticated API boundaries.

Custom Application Logic

MUJPI API Endpoint

Local/Hybrid Inference

Structured Response

Data Flow

Completely controlled by the custom runtime. MUJPI processes inputs statelessly unless explicitly configured to utilize internal RAG or memory stores.

Typical Use Cases

✦Proprietary trading and financial analysis agents
✦Highly regulated healthcare diagnostic assistants
✦Custom robotics and physical automation orchestrators
✦Legacy system integration workflows

Capabilities

✦OpenAPI-compliant inference endpoints
✦High-throughput streaming responses
✦Function calling and structured JSON output
✦Container orchestration support

Security Considerations

Custom runtimes are responsible for their own isolation and security boundaries.
Authentication to MUJPI APIs via strict scoped tokens.
Network policies should restrict MUJPI access solely to the custom runtime IP blocks.

Setup & Deployment

Deployment Modes

Sidecar container deploymentBare-metal adjacent executionRemote API consumption

Setup Overview

Provision an API key with specific model access scopes. Develop your runtime utilizing standard OpenAI-compatible API SDKs pointed at the MUJPI host.