When we started building Shodh Memory, the goal was simple: give AI agents persistent memory. Store things, retrieve things, don't forget between sessions.

But as we used it ourselves, we realized something. Memory isn't enough. A brain doesn't just remember—it organizes, prioritizes, reminds, and plans.

So we kept building. Today, Shodh Memory exposes 37 MCP tools across five categories. It's not just memory anymore. It's a complete cognitive toolkit.

The Five Categories

1. Memory (10 tools)

The core. Persistent, learning memory with Hebbian associations.

Tool	What It Does
`remember`	Store a memory with semantic indexing and entity extraction
`recall`	Semantic search across memories (vector + graph hybrid)
`recall_by_tags`	Filter memories by tags
`recall_by_date`	Find memories within a time range
`forget`	Delete a specific memory
`forget_by_tags`	Bulk delete by tag
`forget_by_date`	Bulk delete by date range
`list_memories`	List all stored memories
`context_summary`	Get a structured summary of decisions, learnings, errors
`proactive_context`	Auto-surface relevant memories for current context

2. Todos (8 tools)

Full GTD (Getting Things Done) workflow. Not a simple checklist—a complete task management system.

Tool	What It Does
`add_todo`	Create a task with project, contexts, priority, due date
`list_todos`	Filter by status, project, context, priority, due date
`update_todo`	Modify any property of a task
`complete_todo`	Mark done (auto-creates next occurrence for recurring tasks)
`delete_todo`	Remove a task
`reorder_todo`	Change priority ordering within a status group
`list_subtasks`	Get child tasks of a parent
`todo_stats`	Counts by status, overdue items, completion rates

Why this matters for AI agents: Claude can now track work across sessions. Start a refactoring task today, continue tomorrow. The agent doesn't forget what needs to be done.

3. Projects (4 tools)

Organize todos into hierarchical projects. Track progress. Archive when done.

Tool	What It Does
`add_project`	Create a project (supports sub-projects via parent)
`list_projects`	See all projects with todo counts and status breakdown
`archive_project`	Hide completed projects (can be restored)
`delete_project`	Permanently remove (optionally with all todos)

4. Reminders (3 tools)

Three trigger types: time-based, duration-based, and context-triggered.

Tool	What It Does
`set_reminder`	Create a reminder with trigger conditions
`list_reminders`	See pending/triggered/dismissed reminders
`dismiss_reminder`	Acknowledge a triggered reminder

Context-triggered reminders are the game-changer:

Context-Triggered Reminderpython

# This reminder surfaces when you mention "deployment"
memory.set_reminder(
    content="Remember to update the changelog before deploying",
    trigger_type="context",
    keywords=["deployment", "deploy", "release", "ship"]
)

# Days later, you say "Let's deploy to production"
# The reminder automatically surfaces: "Remember to update the changelog..."

5. System (7 tools)

Introspection, health checks, and maintenance.

Tool	What It Does
`memory_stats`	Total memories, retrievals, storage size
`verify_index`	Check vector index health, find orphaned memories
`repair_index`	Re-index orphaned memories
`consolidation_report`	See memory strengthening, decay, edge formation
`streaming_status`	Check WebSocket connection health
`token_status`	Monitor context window usage
`reset_token_session`	Reset token counter for new session

Why 37 Tools?

The number isn't the point. The point is completeness.

Most memory solutions give you store/retrieve. That's a database, not a brain.

A brain:

Remembers (memory tools)
Plans and tracks work (todos)
Organizes into domains (projects)
Surfaces things at the right time (reminders)
Has introspection capabilities (system tools)

Shodh Memory provides all of this in a single MCP server. One install, 37 capabilities.

Getting Started

Install Shodh Memory as an MCP server:

Terminalbash

# For Claude Code
claude mcp add shodh-memory -- npx -y @shodh/memory-mcp

Or with Python:

Terminalbash

pip install shodh-memory

All 37 tools become available immediately. No configuration. No cloud setup. Everything runs locally.

The Cognitive Layer

LLMs are powerful reasoners but poor rememberers. They can analyze, synthesize, and generate—but they can't retain.

Shodh Memory is the missing layer: a cognitive substrate that handles memory, planning, and organization so the LLM can focus on reasoning.

37 tools. One MCP server. A complete brain.