Skip to content

component-filtering.md

Latest commit

 

History

History
140 lines (104 loc) · 3.98 KB

component-filtering.md

File metadata and controls

140 lines (104 loc) · 3.98 KB

Component Filtering Configuration

ClojureMCP allows fine-grained control over which tools, prompts, and resources are exposed to AI assistants through configuration options in .clojure-mcp/config.edn. This is useful for creating focused MCP servers with only the components you need.

Table of Contents

Overview

Component filtering uses an allow/deny list pattern:

  • Enable lists (enable-*) - When specified, ONLY these items are enabled
  • Disable lists (disable-*) - Applied after enable filtering to remove specific items
  • Default behavior - When no filtering is specified, all components are enabled

The filtering logic follows this order:

  1. If an enable list is provided and empty ([]), nothing is enabled
  2. If an enable list is provided with items, only those items are enabled
  3. If no enable list is provided (nil), all items start enabled
  4. The disable list is then applied to remove items from the enabled set

Tools Filtering

Control which tools are available to the AI assistant.

Configuration Keys

{:enable-tools [:clojure_eval :read_file :file_write]  ; Only these tools
 :disable-tools [:dispatch_agent :architect]}          ; Remove these tools

Tool Identifiers

Tools can be specified using keywords or strings:

  • :clojure_eval or "clojure_eval"
  • :read_file or "read_file"
  • :file_write or "file_write"

Common tool IDs include:

  • :clojure_eval - Evaluate Clojure code
  • :read_file - Read file contents
  • :file_edit - Edit files
  • :file_write - Write files
  • :bash - Execute shell commands
  • :grep - Search file contents
  • :glob_files - Find files by pattern
  • :dispatch_agent - Launch sub-agents
  • :architect - Technical planning
  • :code_critique - Code review
  • :scratch_pad - Persistent storage

Examples

Minimal REPL-only server:

{:enable-tools [:clojure_eval]}

Read-only exploration server:

{:enable-tools [:read_file :grep :glob_files :LS :clojure_inspect_project]}

Full access except agents:

{:disable-tools [:dispatch_agent :architect :code_critique]}

Prompts Filtering

Control which system prompts are provided to the AI assistant.

Configuration Keys

{:enable-prompts ["clojure_repl_system_prompt" "chat-session-summarize"]
 :disable-prompts ["scratch-pad-save-as"]}

Prompt Names

Prompts are identified by their string names (not keywords):

  • "clojure_repl_system_prompt" - Main REPL interaction prompt
  • "chat-session-summarize" - Session summarization
  • "scratch-pad-load" - Loading scratch pad data
  • "scratch-pad-save-as" - Saving scratch pad snapshots

Examples

Essential prompts only:

{:enable-prompts ["clojure_repl_system_prompt"]}

Disable scratch pad prompts:

{:disable-prompts ["scratch-pad-load" "scratch-pad-save-as"]}

Resources Filtering

Control which resource files are exposed to the AI assistant.

Configuration Keys

{:enable-resources ["PROJECT_SUMMARY.md" "README.md"]
 :disable-resources ["CLAUDE.md" "LLM_CODE_STYLE.md"]}

Resource Names

For now, resources are identified by their string names (not URIs or paths):

  • "PROJECT_SUMMARY.md" - Project overview
  • "README.md" - Main documentation
  • "CLAUDE.md" - Claude-specific instructions
  • "LLM_CODE_STYLE.md" - Coding style guide

Examples

Only project documentation:

{:enable-resources ["PROJECT_SUMMARY.md" "README.md"]}

Remove AI-specific resources:

{:disable-resources ["CLAUDE.md" "LLM_CODE_STYLE.md"]}

See Also