DZ23 Subagents Universal MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CUSTOM_MODEL | Yes | Model ID for the custom provider, e.g. qwen3-coder | |
| DZ23_ROTATION | Yes | Model rotation list, e.g. custom:qwen3-coder | |
| CUSTOM_BASE_URL | Yes | Base URL for the custom OpenAI-compatible server, e.g. http://127.0.0.1:11434/v1 | |
| DZ23_ALLOW_PAID | No | Set to false to block paid and low-cost categories. Categories mixed and free-tier may still incur costs. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| list_modelsC | List configured routing targets, tiers and declared capabilities |
| provider_inventoryB | Inventory all registered providers without exposing secrets |
| discover_modelsC | Discover models from provider /models endpoints |
| health_checkB | Probe configured providers/models in parallel |
| project_initC | Create/update canonical shared project memory |
| mission_statusC | Read shared mission state and recent journal events |
| memory_checkpointC | Persist a handoff checkpoint for another harness/agent |
| delegateC | Delegate one task with automatic provider failover and context continuity |
| consensusC | Ask multiple independent models/reviewers |
| swarm_runC | Run parallel specialist subagents; multiple workers may share a provider |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 10 tools
Tools mostly target distinct resources and actions: provider/model inventory, health, memory, delegation, consensus, and swarm execution. However, list_models and discover_models both return model lists and could be confused, and consensus vs swarm_run both orchestrate multiple models/subagents. Descriptions differentiate them, but boundaries are not perfect.
All names use snake_case, but patterns vary: verb_noun (list_models, discover_models), noun_noun (provider_inventory, health_check, mission_status, memory_checkpoint), noun_verb (project_init, swarm_run), and bare verbs/nouns (delegate, consensus). The inconsistency is readable but not predictable.
10 tools is well within the typical 3-15 range for an orchestration server. Each tool covers a distinct capability area: provider/model management, health, project memory, delegation, consensus, and swarms.
Core workflows for provider/model discovery, health, project memory, mission status, delegation, consensus, and swarm execution are present. Missing or unclear operations include aborting/canceling a running delegate/swarm, dynamically adding/removing providers, and richer mission listing, but these are minor for the apparent scope.