Why SpecMem?¶

The Problem¶

Modern AI coding agents are powerful but have critical limitations:

🧠 Agents Have Amnesia¶

Every session starts fresh. Your agent doesn't remember: - What you built yesterday - Your project's requirements - Previous design decisions - Which tests cover which features

📄 Markdown Chaos¶

AI agents generate mountains of markdown: - requirements.md, design.md, tasks.md - CLAUDE.md, .cursorrules, AGENTS.md

What happens to all these specs after features are built? They sit there, forgotten, while your agent writes code that contradicts them.

🔒 Vendor Lock-In¶

Each agent uses its own format: - Kiro: .kiro/specs/ - Claude Code: CLAUDE.md - Cursor: .cursorrules

Switch agents? Rewrite everything.

⚡ Wasted CI Time¶

Without knowing what changed, agents trigger full test runs for every tiny change.

The Solution¶

SpecMem gives your AI coding agent a memory.

┌─────────────────────────────────────────────────────────────────┐
│                         Your Project                            │
├─────────────────────────────────────────────────────────────────┤
│  .kiro/specs/   CLAUDE.md   .cursorrules   requirements.md      │
│       │             │            │              │                │
│       └─────────────┴────────────┴──────────────┘                │
│                           │                                      │
│                    ┌──────▼──────┐                               │
│                    │   SpecMem   │  ← Unified Memory             │
│                    └──────┬──────┘                               │
│                           │                                      │
│              ┌────────────┼────────────┐                         │
│              │            │            │                         │
│         Dashboard    MCP Server    CLI                           │
│              │            │            │                         │
│           You         Agents      Scripts                        │
└─────────────────────────────────────────────────────────────────┘

Key Benefits¶

1. Persistent Memory¶

Your specs are indexed and searchable. Agents can query them via MCP:

Agent: "What are the authentication requirements?"
SpecMem: Returns relevant specs with context

2. Agent-Agnostic¶

One unified format. Switch from Kiro to Claude Code to Cursor without losing context.

3. Visibility¶

See your project's spec health at a glance: - Which specs are tested? - Which specs are stale? - Which specs have no code references?

4. Smart Testing¶

Know which tests to run when code changes:

specmem impact --files src/auth/service.py
# Shows: auth-requirements.md, auth-design.md
# Suggests: test_auth.py, test_login.py

Who Is SpecMem For?¶

Kiro Users¶

Your .kiro/specs/ become living, searchable memory. The agent knows your requirements when writing code.

Multi-Agent Teams¶

Use Kiro for specs, Claude Code for implementation, Cursor for debugging - all sharing the same context.

Team Leads¶

Dashboard shows project health. See which specs need attention before they become problems.

Quick Comparison¶

Without SpecMem	With SpecMem
Agent forgets specs	Agent queries specs
Specs rot in folders	Specs are living docs
Full test runs	Targeted test runs
Manual context sharing	Automatic context
One agent's format	Any agent's format

Get Started¶

pip install specmem[local]
specmem demo  # See it in action

Then for your project:

cd your-project
specmem init
specmem scan
specmem serve  # Open dashboard

→ Quick Start Guide