oce (Optum Context Engine)

Version: 1.0.0 Description: Claude Code plugin providing MCP tools for the Optum Context Engine (OCE) -- an enterprise RAG platform that retrieves context from 12 internal knowledge sources.

OCE is a context retrieval engine, not a chatbot. It works like Context7 but for Optum/UHG enterprise knowledge: send a query, get back assembled context with citations from HcpDocs, ITSM, GitHub, PRM, and 8 other source systems.

Prerequisites

• python3 (3.10+) available on PATH
• Python packages: httpx, pyyaml, mcp
• OCE API client credentials (OAuth2 client ID and secret)

Setup

1. Install Python Dependencies

pip install -r claude-code/plugins/oce/src/requirements.txt

2. Configure Authentication

Option A -- Environment variables:

export OCE_CLIENT_ID='your-client-id'
export OCE_CLIENT_SECRET='your-client-secret'  # pragma: allowlist secret

Option B -- Config file (~/.claude/oce.local.md):

---
client_id: your-client-id
client_secret: your-client-secret  # pragma: allowlist secret
---

Contact the OCE team for client credentials.

3. Register the MCP Server

Add to your Claude Code config.json (typically ~/.claude/config.json or .claude/config.json):

{
  "mcpServers": {
    "oce-context-engine": {
      "type": "stdio",
      "command": "python3",
      "args": ["claude-code/plugins/oce/src/mcp_server.py"]
    }
  }
}

MCP Tools

Tool	Description
query_context_engine	Query OCE for enterprise knowledge with optional source targeting
list_context_sources	List all 12 available knowledge sources
list_context_configurations	List saved OCE configurations
list_context_models	List available LLM models
get_context_session	Retrieve a previous session's messages and citations

Commands

Command	Description
/oce-admin	Manage OCE applications, configurations, templates, and models

Available Knowledge Sources

Source	Description
hcp-docs	HCP platform documentation and runbooks
hive	Enterprise knowledge base articles
dojo360	Application and service metadata
github	Source code repositories and README files
itsm	ServiceNow incidents, changes, and problems
account-manager	Account and subscription details
infrared	Infrastructure inventory and topology
prm	Problem and risk management records
infrastructure-cost	Cloud spend and cost analysis
techlandscape	Technology governance and standards
techops-hub	Operational procedures and playbooks
kubernetes	Cluster, namespace, and workload information

Usage Examples

Basic enterprise query

Claude automatically uses OCE when you ask about enterprise topics:

"How do I configure Azure Private DNS zones for HCP?"

Targeted source query

Specify sources for more focused results:

"Search HCP docs and TechOpsHub for the VM patching runbook"

This maps to:

query_context_engine(
    query="VM patching runbook",
    sources=["hcp-docs", "techops-hub"]
)

Follow-up queries

Use session context for deeper exploration:

# First query
query_context_engine(query="What is the HCP backup strategy?")
# Returns session_id: "abc-123"

# Follow-up in same context
query_context_engine(
    query="What about disaster recovery?",
    session_id="abc-123"
)

Administrative tasks

/oce-admin status
/oce-admin models list
/oce-admin configs list
/oce-admin sources

Architecture

Claude Code
    |
    v
MCP Server (mcp_server.py)  <-- stdio transport
    |
    v
OCE Client (oce_client.py)  <-- OAuth2 + httpx
    |
    v
OCE API (api.uhg.com)       <-- RAG platform
    |
    v
12 Knowledge Sources         <-- HcpDocs, ITSM, GitHub, ...

The MCP server translates tool calls into OCE API requests. The sources parameter on query_context_engine is mapped to OCE's chatSettings agent/plugin configuration to scope queries to relevant knowledge sources.