What is TekMemo?

TekMemo is a layered memory runtime for AI agents and applications.

In a world where LLMs have large but ephemeral context windows, TekMemo provides the "long-term memory" and "project grounding" needed to keep agents accurate, consistent, and context-aware over weeks or months of work.

Why TekMemo?

Current AI tools often suffer from "context drift" or "knowledge gaps" because they lack a dedicated place to store and retrieve project-specific knowledge. TekMemo solves this by providing:

1. Grounded Context: Instead of relying on the model's training data, you ground it in the specific decisions, requirements, and architecture of your project.
2. Durable Memory: Notes, decisions, and patterns are stored in a structured way that persists across sessions.
3. File-First Design: Memory is stored as plain text and JSON files in your project's .tekmemo folder. It's human-readable, version-controllable, and easily inspectable.
4. Layered Approach: Memory is organized into distinct layers:
   • Core: The essential "project briefing" that is always relevant.
   • Notes: Long-form archival records, decisions, and summaries.
   • Recall: Indexed fragments for fast semantic retrieval.
   • Graph: Structural relationships and architectural dependencies.

How it Works

TekMemo works as a bridge between your project and your AI agent:

• Collection: Agents or developers record facts, decisions, and summaries into the .tekmemo directory.
• Organization: The runtime automatically manages snapshots, indexing, and conflict resolution.
• Retrieval: When an agent starts a task, TekMemo performs "Recall" (semantic search) and "Composition" to inject the most relevant memory directly into the agent's prompt.

Getting Started

This guide gets you from zero to a working TekMemo memory setup.

Recommended path

1. Install TekMemo
2. Initialize local memory
3. Use the Tekmemo client
4. Understand the filesystem
5. Add durable records
6. Use the CLI in an agent workflow
7. Use the MCP in an agent workflow
8. Connect cloud when you need sync or hosted APIs

Runtime choices

All runtime modes are selected when constructing a Tekmemo client (or the equivalent CLI/MCP flag). The client is backed by a strategy per mode:

Runtime	Use when
Local	You want inspectable project memory stored in .tekmemo/.
Cloud	You want hosted project memory, sync, API keys, recall, and graph endpoints.
Hybrid	You want local files plus cloud recall/sync, routed by read/write policies.
In-memory	You are writing tests, demos, or examples. Nothing persists.

See Configuration for the resolution chain and The Tekmemo client for the full API surface.