get_instructions

Instructions

Implementation Reference

• src/server/tools/CoreTools.ts:17-26 (schema)

  Tool definition/schema for 'get_instructions' - defines the tool name, description, and empty inputSchema (no parameters required). The description instructs users to call this FIRST at the start of every session.

  export const coreTools = [
    {
      name: 'get_instructions',
      description: '⚠️ CALL THIS FIRST. Get comprehensive instructions for using the Memory Bank MCP server. Call this tool FIRST at the start of every session to understand the available tools and recommended workflow.',
      inputSchema: {
        type: 'object',
        properties: {},
        additionalProperties: false,
      },
    },

• src/server/tools/CoreTools.ts:378-523 (handler)

  The handler function `handleGetInstructions()` - returns static comprehensive Markdown instructions for using the Memory Bank MCP server. No Memory Bank initialization required. Contains the full tool reference, workflow, valid modes, core files, and quick-start checklist.

  export function handleGetInstructions() {
    const instructions = `# Memory Bank MCP Server — Instructions
  
  ## What is Memory Bank?
  Memory Bank is an MCP server that persists project context across AI sessions.
  It stores progress, decisions, tasks, and a knowledge graph in a \`memory-bank/\`
  folder inside your project so that every new session can pick up where the last
  one left off.
  
  ## Mandatory Workflow (every task, no exceptions)
  
  ### START of task
  1. Call \`get_instructions\` (this tool) — you only need to do this once per session.
  2. Call \`get_context_digest\` to load the compact summary of current project state
     (tasks, issues, recent progress, recent decisions, and knowledge graph summary).
  3. Call \`graph_search\` with a relevant query to find knowledge graph entities
     related to the current task.
  
  ### DURING task
  4. Call \`track_progress\` after completing milestones.
  5. Call \`log_decision\` when making architectural or design choices.
  6. Call \`add_session_note\` for observations, blockers, or questions.
  
  ### END of task
  7. Call \`update_active_context\` with updated tasks, issues, and next steps.
  8. Call \`track_progress\` with a final summary of what was accomplished.
  9. Update knowledge graph entities if the project structure changed
     (use \`graph_upsert_entity\`, \`graph_add_observation\`, \`graph_link_entities\`).
  
  ## If Memory Bank Contains Placeholder Text
  If any core file contains \`[Project description]\` or \`[Task 1]\` style placeholders,
  the Memory Bank has never been populated. You MUST fill all core files with real
  project data before doing any other work.
  
  ## Valid Modes
  The ONLY valid modes are: architect, code, ask, debug, test.
  There is NO "full" mode. All tools are available in every mode — modes control
  behavior guidelines, not tool access. Use \`switch_mode\` to change modes.
  
  ## Core Memory Bank Files
  These files live in the \`memory-bank/\` directory:
  - **product-context.md** — Project overview, objectives, technologies, architecture
  - **active-context.md** — Ongoing tasks, known issues, next steps, session notes
  - **progress.md** — Completed & pending milestones, update history
  - **decision-log.md** — Architectural/design decisions with context and alternatives
  - **system-patterns.md** — Architecture, code, and documentation patterns.
    **IMPORTANT:** Read this file when starting work to understand project conventions.
    Update it whenever you introduce new patterns, change architecture, or adopt new coding conventions.
  
  ## Complete Tool Reference
  
  ### 1. Instructions
  | Tool | Purpose |
  |------|---------|
  | \`get_instructions\` | Return these instructions (call once per session) |
  
  ### 2. Context Loading (read-only)
  | Tool | Purpose |
  |------|---------|
  | \`get_context_digest\` | Compact summary: recent progress, tasks, issues, decisions, graph overview |
  | \`get_context_bundle\` | Full content of ALL core files in one response (larger payload) |
  | \`get_memory_bank_status\` | Status of the Memory Bank (initialized, path, file list) |
  | \`list_memory_bank_files\` | List all files in the Memory Bank directory |
  | \`search_memory_bank\` | Full-text search across all Memory Bank files |
  
  ### 3. File Operations
  | Tool | Purpose |
  |------|---------|
  | \`read_memory_bank_file\` | Read single file. **Params:** \`filename\` (string) |
  | \`write_memory_bank_file\` | Write single file. **Params:** \`filename\`, \`content\` (strings), optional \`ifMatchEtag\` |
  | \`batch_read_files\` | Read multiple files. **Params:** \`files\` (string array, e.g., ["progress.md", "active-context.md"]) |
  | \`batch_write_files\` | Write multiple files. **Params:** \`files\` (array of {filename, content, ifMatchEtag?}) |
  
  ### 4. Progress Tracking
  | Tool | Purpose |
  |------|---------|
  | \`track_progress\` | Record a progress milestone. **Params:** \`action\` (string), \`description\` (string) |
  | \`add_progress_entry\` | Structured progress entry. **Params:** \`type\` (feature/fix/refactor/docs/test/chore), \`summary\`, \`details\`, \`files\`, \`tags\` |
  | \`update_active_context\` | Update active-context.md. **Params:** \`tasks\`, \`issues\`, \`nextSteps\` (all string arrays) |
  | \`update_tasks\` | Modify tasks list. **Params:** \`add\`, \`remove\`, or \`replace\` (string arrays) |
  | \`add_session_note\` | Add timestamped note. **Params:** \`note\` (string), \`category\` (observation/blocker/question/decision/todo) |
  | \`log_decision\` | Log a decision. **Params:** \`title\`, \`context\`, \`decision\` (all strings) |
  
  ### 5. Knowledge Graph
  | Tool | Purpose |
  |------|---------|
  | \`get_targeted_context\` | **Preferred.** Budgeted context via KG + excerpts. **Params:** \`query\` (string), optional \`budget\`, \`entities\` |
  | \`graph_search\` | Search entities. **Params:** \`query\` (string), optional \`limit\` |
  | \`graph_open_nodes\` | Expand nodes and neighbors. **Params:** \`nodes\` (string array), optional \`depth\` |
  | \`graph_upsert_entity\` | Create/update entity. **Params:** \`name\`, \`entityType\` (strings), optional \`attrs\` |
  | \`graph_add_observation\` | Add observation. **Params:** \`entity\` (name string), \`text\` (string) |
  | \`graph_add_doc_pointer\` | Link entity to file. **Params:** \`entity\`, \`docPath\`, optional \`heading\` |
  | \`graph_link_entities\` | Create relationship. **Params:** \`from\`, \`to\`, \`relationType\` (all strings) |
  | \`graph_delete_entity\` | Delete entity/observation. **Params:** \`entity\` (name string), optional \`observationId\` |
  | \`graph_maintain\` | Maintenance ops. **Params:** \`operation\` (rebuild/compact/stats) |
  
  ### 6. Modes
  | Tool | Purpose |
  |------|---------|
  | \`switch_mode\` | Switch to a mode: architect, code, ask, debug, or test |
  
  ### 7. Setup & Administration
  | Tool | Purpose |
  |------|---------|
  | \`initialize_memory_bank\` | Create a new Memory Bank in a directory |
  | \`set_memory_bank_path\` | Point to an existing Memory Bank directory |
  | \`migrate_file_naming\` | Migrate files from camelCase to kebab-case |
  | \`debug_mcp_config\` | Debug the current MCP configuration |
  
  ### 8. Backup & Restore
  | Tool | Purpose |
  |------|---------|
  | \`create_backup\` | Create a timestamped backup (use listOnly:true to list backups) |
  | \`restore_backup\` | Restore from a backup (auto-creates pre-restore backup) |
  
  ### 9. Multi-Store Management
  | Tool | Purpose |
  |------|---------|
  | \`list_stores\` | List all registered stores (use action:"register"/"unregister" to manage) |
  | \`select_store\` | Switch the active store (by path or storeId) |
  
  ### 10. Sequential Thinking
  | Tool | Purpose |
  |------|---------|
  | \`sequential_thinking\` | Record thinking step. **Params:** \`thought\`, \`thoughtNumber\`, \`totalThoughts\` (int), \`nextThoughtNeeded\` (bool). Use \`reset:true\` to clear |
  | \`finalize_thinking_session\` | Persist outcomes. **Params:** \`summary\` (string), optional \`decisions\`, \`tasks\`, \`progress\` |
  
  ## Quick-Start Checklist
  1. \`get_instructions\` — read these instructions (done!)
  2. \`get_context_digest\` — load current project state
  3. \`get_targeted_context\` or \`graph_search\` — find relevant context for your task
  4. Do your work, calling \`track_progress\` / \`log_decision\` / \`add_session_note\` as you go
  5. Use \`sequential_thinking\` for complex reasoning, then \`finalize_thinking_session\` to persist
  6. \`update_active_context\` — save updated tasks, issues, next steps
  7. \`track_progress\` — final summary of accomplishments
  `;
  
    return {
      content: [
        {
          type: 'text',
          text: instructions,
        },
      ],
    };
  }

• src/server/tools/index.ts:82-86 (registration)

  Registration of the 'get_instructions' tool handler in the switch statement within `setupToolHandlers`. When the tool name is 'get_instructions', it calls `handleGetInstructions()` directly without needing Memory Bank.

  switch (request.params.name) {
    // Instructions (no Memory Bank required)
    case 'get_instructions': {
      return handleGetInstructions();
    }

• src/server/tools/index.ts:11-11 (helper)

  Import of `handleGetInstructions` from CoreTools.ts, making it available for registration in the tool handler switch statement.

  import { coreTools, handleGetInstructions, handleSetMemoryBankPath, handleInitializeMemoryBank, handleReadMemoryBankFile, handleWriteMemoryBankFile, handleListMemoryBankFiles, handleGetMemoryBankStatus, handleMigrateFileNaming, handleDebugMcpConfig, handleGetContextBundle, handleGetContextDigest, handleSearchMemoryBank, handleCreateBackup, handleRestoreBackup, handleAddProgressEntry, handleAddSessionNote, handleUpdateTasks, handleBatchReadFiles, handleBatchWriteFiles } from './CoreTools.js';