★ Welcome to RECALL! ★

RECALL is a local memory layer for AI agents. It helps your AI assistant remember a project across sessions by storing notes, imports, decisions, and handoffs in a local knowledge base — instead of relying on one chat window or one model's built-in memory.

C:\> npm install -g @recallmeridian/recall

What Is RECALL?

RECALL gives AI agents a durable, local memory for real projects. It captures project context, keeps uncertain material separate from trusted knowledge, and gives Claude, ChatGPT, and other agents a safer way to search, continue, audit, and build on previous work.

📁

Knowledge Base

Per-project entries for decisions, milestones, bugs, and architecture. Versioned and queryable.

 💾

Session Handoff

Type /recall in any AI session. Get the full project state in under 300 words.

 🖥️

Local Dashboard

Self-contained HTML dashboard. Runs on file://. No server required.

Why RECALL?

Slash commands are great. But they aren't memory.

"Slash commands tell an agent what to do next. RECALL helps the agent remember what matters, know what is trusted, and continue work without starting over." — the RECALL positioning, 2026
Slash Commands RECALL
✗ Trigger a prompt or workflow ✓ Stores durable project memory
✗ Session-bound ✓ Works across sessions and tools
✗ Depends on what the AI remembers ✓ Reads from a local knowledge base
✗ "Do this task" ✓ "Understand the project, help me continue"

How RECALL Is Different

Most existing AI memory tools help one model remember preferences or keep context inside one product. RECALL is different because it is local, project-shaped, auditable, and tool-agnostic.

vs. ChatGPT Memory & Projects

ChatGPT Memory and Projects help ChatGPT personalize responses and keep project context inside ChatGPT. OpenAI describes project memory as drawing context from conversations within a project. RECALL is not just conversation memory — it's a local project knowledge system with draft/trusted states, import workflows, ledgers, and review gates.

Sources: ChatGPT Memory | ChatGPT Projects

vs. Cursor Rules & Memories

Cursor Rules and Memories help coding agents keep reusable context or project instructions inside one IDE. RECALL stores structured project knowledge, supports evidence promotion, tracks handoffs, and serves multiple agents — not just one IDE.

Sources: Cursor Rules | Cursor Memories

RECALL Is Not...

  • a prompt library
  • a chat memory feature
  • a notes app
  • an IDE rule file
  • a vector database

RECALL is closer to a local operating memory for AI-assisted work: searchable, reviewable, project-aware, and designed to keep long-running work coherent.

How It Works

Three layers. Tiny context. Big memory.

Layer What Size
L1 Memory pointer ~50 tokens
L2 KB summary ~3 KB
L3 Full JSON On demand

Most "memory" tools dump everything into context and hope for the best. RECALL starts with a 50-token pointer, expands to a 3KB summary when needed, and only loads the full JSON when you ask for it.

Slash Commands

Twelve commands installed. Five most useful below:

/recall » Project handoff — milestones, TODOs, next step
/recall-sync » Pull latest session data into the dashboard
/recall-kb » Add, list, or update KB entries by category
/recall-milestone » Mark a milestone complete or queue next
/recall-analyze » Run AI analysis over recent sessions
+ 7 more — see GitHub for full reference

Not ready for the waitlist? Help shape what RECALL becomes — take our 2-minute feedback survey »

~ Thanks for visiting! Come back soon! ~