OpenClaw Integration Guide

Give your OpenClaw personal assistant a persistent memory layer with AgentRecall. The assistant remembers user preferences, past conversations, and context across every channel.

Overview

OpenClaw is a personal assistant platform built on the Model Context Protocol (MCP). By adding the AgentRecall MCP server, your assistant gains the ability to store and retrieve memories — facts, preferences, corrections, and temporal context — across sessions and channels.

Once configured, the agent automatically indexes important information from conversations and surfaces relevant memories in future interactions. No prompt engineering required.

Prerequisites

  1. Python 3.10 or later installed on your system.
  2. OpenClaw installed and functional (openclaw --version should work).
  3. An AgentRecall account with an active API key (ark_...).

Installation

Install the AgentRecall MCP server package globally:

pip install agentrecall-mcp
Tip
If you use a virtual environment for OpenClaw, make sure to install the package inside that environment so the binary is on the PATH when OpenClaw launches the MCP server.

Setup

Register the AgentRecall MCP server with OpenClaw in a single command:

openclaw mcp set agentrecall "{\n  "command": "agentrecall-mcp",\n  "env": {\n    "AGENTRECALL_API_KEY": "ark_YOUR_KEY",\n    "AGENTRECALL_AGENT_ID": "openclaw"\n  }\n}"

Replace ark_YOUR_KEY with the API key from your AgentRecall dashboard. The AGENTRECALL_AGENT_ID should match the agent name you want memories scoped to.

Verify Installation

Confirm the server is registered:

openclaw mcp list

You should see `agentrecall` in the output. To view full details:

openclaw mcp show agentrecall

If the server appears and its status shows healthy, you are ready. The agent will begin persisting memories automatically.

Configuration Reference

All configuration is passed through environment variables when the MCP server is launched. Adjust them in the mcp set JSON.

VariableDescriptionRequiredDefault
AGENTRECALL_API_KEYYour AgentRecall API key (ark_...)Yes—
AGENTRECALL_AGENT_IDAgent identifier (e.g. openclaw)Yeshermes
AGENTRECALL_BASE_URLCustom API endpointNohttps://api.agentrecall.com
AGENTRECALL_LOG_LEVELLogging verbosityNoinfo

Use Cases

User Preferences

The assistant learns that you prefer dark mode, concise responses, and use VS Code — and applies these preferences automatically in every session.

Relationship Memory

Store context about the people in your life — birthdays, communication styles, preferences — so the assistant can help you plan and remember.

Daily Routines

Track recurring tasks, schedules, and habits. The assistant can surface reminders and suggest routines based on past behavior.

Cross-Channel Context

Whether you interact via CLI, Telegram, or Slack, the same memory store ensures continuity. Conversations started on one platform carry over to another.

Multi-Agent Setup

OpenClaw can run multiple agents simultaneously. You can register separate AgentRecall agents for different purposes:

{\n  "personal": {\n    "env": {\n      "AGENTRECALL_AGENT_ID": "openclaw-personal"\n    }\n  },\n  "coding": {\n    "env": {\n      "AGENTRECALL_AGENT_ID": "openclaw-coding"\n    }\n  }\n}

Each agent maintains its own isolated memory graph. Personal memories won't leak into coding context and vice versa.

Troubleshooting

Server not showing in mcp list
Ensure the agentrecall-mcp binary is on your PATH. Run which agentrecall-mcp to verify. If not found, reinstall with pip install --force-reinstall agentrecall-mcp.
Authentication errors
Double-check your API key starts with ark_ and has not expired. You can verify it on the AgentRecall dashboard under API Keys.
Memories not appearing
The server stores memories asynchronously. Give it a few seconds after the relevant conversation. You can also search directly via the dashboard to confirm.

Removing the Integration

To completely remove AgentRecall from your OpenClaw setup:

openclaw mcp unset agentrecall

This disconnects the server. Stored memories remain in your AgentRecall account and can be accessed from the dashboard or reconnected later.