How beehiveinnovations/zen-mcp-server works
PAL MCP is a Model Context Protocol server that lets an AI client invoke specialized coding tools and consult multiple model providers while maintaining conversation context. The primary path is client request → server tool dispatch → simple or guided workflow execution → model response returned through the server; conversation state supports continuity across turns. The README also documents multi-pass review, planning, debugging and other coding workflows, plus clink’s external CLI bridge. Sampled source confirms server dispatch, shared tool/workflow infrastructure, and provider registries; unsampled tool wiring is not expanded.
Main entry points: server.py MCP server · providers Provider adapters · clink.py Clink tool · base.py Simple tool runtime
Components 18
MCP entry
- MCP server
server.py
Coding tools
- Simple tool runtime
tools/simple/base.py - Workflow runtime
tools/workflow/workflow_mixin.py - Code review
tools/codereview.py - Coding workflows
tools - Workflow prompts
systemprompts
Conversation context
- Conversation memory
utils/conversation_memory.py - Thread storage
utils/storage_backend.py
Model access
- Provider registry
providers/registry.py - Provider adapters
providers - Model registries
providers/registries - Model capabilities
providers/shared/model_capabilities.py
CLI bridge
- Clink tool
tools/clink.py - CLI registry
clink/registry.py - CLI agents
clink/agents
Other components
- AI client
- Model services
- External CLI
Connections 20
- AI client → MCP server: requests tools
- MCP server → AI client: returns results
- MCP server → Simple tool runtime: dispatches
- MCP server → Workflow runtime: dispatches
- MCP server → Conversation memory: manages turns
- Simple tool runtime → Provider registry: requests model
- Workflow runtime → Conversation memory: reads thread
- Conversation memory → Thread storage: reads and writes
- Provider registry → Provider adapters: selects provider
- Provider adapters → Model registries: resolves models
- Model registries → Model capabilities: provides capabilities
- Provider adapters → Model services: sends requests
- Model services → Provider adapters: returns responses
- AI client → Clink tool: invokes bridge
- Clink tool → CLI registry: loads CLI config
- Clink tool → CLI agents: launches agent
- CLI agents → External CLI: runs CLI
- Code review → Workflow runtime: uses workflow runtime
- Code review → Workflow prompts: uses review prompt
- Coding workflows → Simple tool runtime: includes tools