Palaia

Your agent forgets everything. Every. Single. Session.

Yesterday's decisions, last week's server IP, that config you spent 20 minutes debugging — gone. Your agent starts over from scratch, makes the same mistakes, asks the same questions.

Palaia gives your agent a memory that actually works. Local. Private. Instant.

Get Started

Tell your agent:

"Install or update the Palaia memory skill from ClawHub to the latest version (even if already present). Read the SKILL.md completely and follow it step by step. Run palaia init, then palaia doctor --fix and resolve all warnings — don't stop until the doctor report is clean. Set up completely."

That's it. Your agent handles everything. You just confirm.

No cloud. No API keys. No database server. Everything stays on your machine.

Why Palaia?

Because the alternatives suck.

Cloud memory services need API keys, internet, and trust. Vector databases need infrastructure. Dumping everything into markdown files works until you have 200 of them and can't find anything.

Palaia runs on your machine, survives crashes (WAL-backed), and finds things by meaning — not just keywords. Search for "due date" and it finds "the deadline is March 15th."

Core Concepts

Write, Search, Find

# Save something worth remembering
palaia write "Customer prefers CSV exports over PDF" --tags "preferences"

# Find it weeks later
palaia query "what format does the customer want"

# See what's in active memory
palaia list

Projects

Keep memories organized when juggling multiple tasks:

palaia project create website-redesign --description "Q2 redesign"
palaia project write website-redesign "Homepage must load under 2s"
palaia project query website-redesign "performance targets"

Entry Types

Classify what you store:

palaia write "Always run migrations before deploy" --type process --tags "deploy"
palaia write "Switch to FastAPI decided 2026-03-01" --type memory --tags "decision,adr"
palaia write "Fix login timeout bug" --type task --status open --priority high

Scopes

Control visibility:

• private — Only the agent that wrote it
• team — All agents in the workspace (default)
• public — Exportable across workspaces

Smart Tiering

Palaia auto-manages memory over time:

• HOT — Frequently accessed, always in search results
• WARM — Untouched for ~7 days, still searchable
• COLD — Untouched for ~30 days, archived but retrievable

Nothing gets deleted. Old memories fade quietly to the background.

Semantic Search

Palaia finds things by meaning, not just keywords. Available providers:

palaia detect                                    # See what's available
palaia config set-chain fastembed bm25           # Set provider priority
palaia warmup                                    # Pre-build search index

Multi-Agent Setup

Multiple agents can share one memory store:

palaia init                          # Auto-detects agents
palaia write "note" --agent elliot   # Attribute to specific agent
palaia write "secret" --scope private  # Only visible to the writer

Document Ingestion (RAG)

Index external documents alongside agent memory:

palaia ingest document.pdf --project docs
palaia ingest https://example.com/api.html --project api-docs
palaia query "How does auth work?" --project api-docs --rag

CLI Reference

All commands support --json for machine-readable output.

OpenClaw Plugin

Replace OpenClaw's built-in memory with Palaia:

npm install @byte5ai/palaia

Add to your OpenClaw config:

{ "plugins": ["@byte5ai/palaia"] }

Manual Install (without OpenClaw)

pip install "palaia[fastembed]"    # or: uv tool install "palaia[fastembed]"
palaia init
palaia doctor --fix
palaia warmup                      # pre-builds search index — don't skip this

Development

git clone https://github.com/iret77/palaia.git
cd palaia
pip install -e ".[dev]"
pytest

550+ tests. Contributions welcome.

Links

• CHANGELOG — What's new in each version
• ClawHub — Install via agent
• GitHub — Source + Issues
• OpenClaw — The agent platform Palaia is built for

MIT — © 2026 byte5 GmbH