Skip to main content

MemClaw Setup: Connect Claude Code to Persistent Project Workspaces

· 2 min read
Felo AI
Felo AI
Operations

Step-by-step guide to connecting Claude Code to your MemClaw workspace via MCP. Configure mcp_config.json, test the connection, and set up slash commands in 5 minutes.

If you've been searching for how to give Claude Code persistent memory, you've probably come across various memory server options. MemClaw is a different kind of solution — a project workspace manager, not a global memory store.

MemClaw vs memory servers comparison

MemClaw Is a Workspace Manager​

Global memory servers store facts across all your conversations. MemClaw gives each project its own isolated workspace — context, tasks, artifacts, and decisions that are scoped to that project only. When you open a project, the agent loads that workspace and knows exactly where things left off.

The key difference: global memory shares context across everything. MemClaw isolates context per project — zero cross-project bleed by design.

MemClaw workspace overview

What MemClaw Stores Per Project​

  • Living README — project background, preferences, constraints
  • Artifacts — documents, reports, URLs, files
  • Tasks — auto-tracked as the agent works
  • Decisions — why you chose X, logged permanently

Context restore: 8 seconds. No re-briefing, no copy-pasting.

Setup​

How to Install MemClaw​

Claude Code Plugin Marketplace (recommended):

/plugin marketplace add Felo-Inc/memclaw
/plugin install memclaw@memclaw
export FELO_API_KEY="your-api-key-here"

Get your API key at felo.ai/settings/api-keys. OpenClaw:

bash <(curl -s https://raw.githubusercontent.com/Felo-Inc/memclaw/main/scripts/openclaw-install.sh)

Using Your Workspace​

Once installed, open any project and load its workspace:

load workspace my-project

The agent restores full context in about 8 seconds. From there, it knows what's in progress, what's been decided, and what still needs doing.

To save something explicitly:

save decision: using Postgres not MySQL — JSON query requirements
save artifact: client-brief.pdf
mark task complete: auth module

Tasks get tracked automatically as the agent works.

Who This Is For​

MemClaw works for anyone managing multiple projects with AI agents — not just developers:

  • Freelancers with multiple clients — one workspace per client, zero bleed
  • Product managers tracking multiple features in parallel
  • Sales professionals managing multiple accounts with different histories
  • Developers who need persistent codebase context between sessions

No JSON config. No server process. One command to install.

Get started at memclaw.me — GitHub: github.com/Felo-Inc/memclaw