MCP-native · 12 memory tools

MemClaw — the shared, governedMCP memory server for AI agents.

Persistent memory over one MCP connection. What one agent learns, every agent knows — in Claude Code, Claude Desktop, Cursor, Windsurf, and every framework that speaks MCP.

Get Started Free →

No credit card

Open-Source Repo

Apache 2.0 · 70,000 downloads

The entire setup — this is all of it

{
  "mcpServers": {
    "memclaw": {
      "url": "https://memclaw.net/mcp",
      "headers": {
        "X-API-Key": "mc_your_key"
      }
    }
  }
}

Paste into any MCP client — 12 memory tools appear

Connected in 30 seconds

Three steps. The longest one is typing your email.

1~30 seconds

Create a free account

Email or SSO. No credit card, no sales call.

2one click

Copy your API key

A tenant-scoped key (starts with mc_) is waiting on your dashboard.

3done

Paste the config

Claude Code, Claude Desktop, Cursor, Windsurf — 12 memory tools appear instantly.

The 12 tools your agents get

Not a key-value cache. Writes are auto-enriched, recall is semantic, and every operation is scoped, attributed, and audited.

memclaw_write

Save a fact once — auto-enriched with type, title, summary, and tags.

memclaw_recall

Semantic search over everything your agents know.

memclaw_list

Browse what's stored, newest first, no query needed.

memclaw_manage

Update, supersede, or delete memories as facts change.

memclaw_evolve

Record outcomes so stored knowledge learns from results.

memclaw_insights

Patterns, contradictions, and gaps across the store.

memclaw_tune

Retrieval knobs per search profile — depth, hops, thresholds.

memclaw_doc

Store and fetch whole documents and skills.

memclaw_entity_get

The graph view of a person, project, or system.

memclaw_keystones

Read the fleet's mandatory policy rules.

memclaw_keystones_set

Author fleet-wide policy (elevated trust required).

memclaw_stats

Usage, quotas, and health for your tenant.

Memory for your agent framework

MCP clients connect with the config above. Framework SDKs get their own guide — same memory, same governance, either way.

Claude CodeIntegration guide →Claude DesktopIntegration guide →CursorIntegration guide →OpenClawIntegration guide →LangChainIntegration guide →LlamaIndexIntegration guide →CrewAIIntegration guide →AutoGenIntegration guide →OpenAI Agents SDKIntegration guide →

Something else? Generic MCP setup · Plain REST API

Where to go from here

Everything on this site, one hop away.

Learn

Evaluate

Run

Already have an account? Open your dashboard

Frequently asked questions

What is an MCP memory server?+

MCP (Model Context Protocol) is the open standard AI clients use to call external tools. An MCP memory server exposes memory as those tools — so Claude Code, Claude Desktop, Cursor, or any MCP client can write facts, recall them semantically, and share them with other agents. MemClaw adds what teams need on top: multi-agent scopes, fleets, policy rules, and an audit trail.

Which AI clients and frameworks work with it?+

Anything that speaks MCP connects with the one config on this page — Claude Code, Claude Desktop, Cursor, and Windsurf included. Framework SDKs (LangChain, LlamaIndex, CrewAI, AutoGen, OpenAI Agents) have their own integration guides, and there's a plain REST API when MCP isn't an option.

Is it open source? Can I self-host?+

Yes — caura-memclaw is Apache 2.0 on GitHub: the whole engine, including all 12 MCP tools and the audit trail. The managed platform at memclaw.net is the same engine with governance, SSO, and hosting wired in, and its free tier needs no credit card.

One config. Every agent remembers.

Free tier: 10K memories, unlimited agents and fleets. No credit card.

Get Started Free →

No credit card

Open-Source Repo

Apache 2.0 · 70,000 downloads