Guide · Agent skill

How to use the agentmemory Agent Connector agent skill

Updated 2026-10-29 · written with AI from the project's own documentation and checked against our catalogue data

The agentmemory-agents skill teaches an AI agent how to connect the agentmemory memory server to a host coding agent such as Claude Code, Cursor or Codex. It runs the agentmemory connect command, which merges the server into the host's config and keeps any servers already listed there. This guide explains what the skill does, how to install it, how to trigger it, and what to check before and after it changes your setup.

agentmemory Agent Connectorrohitg00/agentmemory · Apache-2.0 Download the skill

Quick start

1Install agentmemory CLI. Make sure the agentmemory CLI is installed and available on your command line before you start.
2Install the skill. Unzip the downloaded ZIP into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills.
3Connect your host agent. Run agentmemory connect with your host's name, such as claude-code, and review the result.
4Restart or reload the host. Restart the host or run its MCP reload, for example /mcp in Claude Code, so it picks up the server.
5Confirm the tools appear. Check that the agent lists agentmemory's tools in its tool inventory.
6Run a save and recall check. Ask the agent to save a short test memory and then recall it to confirm the connection works.

What the agentmemory-agents skill does

Agent skills are instruction packages that an AI agent loads when a request matches them. This one covers a single job: wiring agentmemory into a host coding agent. Its description says it applies when you install agentmemory into a specific agent, when you ask which agents are supported, or when a connector writes the wrong config path.

The skill tells the agent to run agentmemory connect with the name of the host, then restart or reload that host, confirm the agentmemory tools appear, and run a save and recall check. For some hosts it also lists extra steps, such as Pi, Cursor or Devin CLI.

Detects which agent is calling and asks which host to configure if it is unknown
Runs agentmemory connect <name> with a name from the adapter table
Checks that the host lists agentmemory's tools after a restart or reload
Performs an authorized save and recall check

Install the skill from the ZIP

Download the ZIP from this page and unzip it. For a personal installation, place the unzipped skill folder in ~/.claude/skills/. To limit it to one project, place it in that project's .claude/skills/ folder.

In the Claude apps you can instead upload the ZIP under Settings → Capabilities → Skills.

Personal: ~/.claude/skills/
Single project: .claude/skills/ inside the project
Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Install the agentmemory CLI and connect your agent

The skill needs the agentmemory CLI installed and available on the command line. Once it is, connect your host with the matching name. The source shows claude-code, cursor, codex and gemini-cli as examples; the full list of 22 adapter names is in REFERENCE.md in the skill folder.

After wiring, restart the host or run its MCP reload. In Claude Code that is /mcp. Then confirm the agent lists agentmemory's tools.

agentmemory connect claude-code # or cursor, codex, gemini-cli, ...

Connect Pi, Oh My Pi or DeepSeek Harness

These three hosts need a running agentmemory daemon first. Start it, then run the matching command. Pi and OMP install native recall, capture and bundled skills, and --with-mcp adds the tools needed for broader workflows.

After connecting Pi, reload it with /reload, or restart OMP, then run /agentmemory-status. Restart DSH after enabling its bundled capture plugin. Re-run the matching connector after upgrading or moving agentmemory so installed runtime paths stay current.

agentmemory connect pi --with-mcp
agentmemory connect omp --with-mcp
agentmemory connect dsh --with-hooks

Set up Cursor and Devin CLI

Cursor's connector configures MCP. For lifecycle capture, install and enable the Cursor plugin through Settings → Plugins → Add marketplace, then add this repository. Do not add a second global hook copy.

Devin CLI and Devin Desktop can install the whole plugin in one step, which adds the skills, the MCP server and the capture hooks. Cloud Devin sessions cannot reach a local agentmemory server, so use the plugin only for local sessions.

Cursor: connect for MCP, then add the plugin for lifecycle capture
Devin CLI: plugin install adds skills, MCP server and capture hooks
Cloud Devin sessions: not supported, because they cannot reach a local server

devin plugins install rohitg00/agentmemory#plugin

Install skills separately for Hermes or OpenClaw

Hermes, OpenClaw, Pi and OMP connectors bundle skills already. Only use a separate skills install when you need one. Choose the intended host, and add --global only for a user-wide installation. Avoid duplicate installs when a native package already includes the skills.

Native Windows supports agentmemory connect copilot-cli. Other automated connectors need macOS, Linux, or a host running in the same WSL2 environment. Otherwise, follow the host's manual setup instructions.

Requires npx and the skills package version 1.7.1
Hermes example: --agent hermes-agent
OpenClaw example: --agent openclaw

npx skills@1.7.1 add rohitg00/agentmemory --agent hermes-agent --skill '*'

Troubleshoot a connector or missing tools

If a connector prints manual setup instructions, installation is not finished. Restart the configured host, then check its actual tool inventory. A native memory provider and an MCP connection are separate, so a working native save or search does not prove the MCP tools are connected.

If the connector wrote the wrong config path, do not run another host's setup commands. Check the adapter's entry in REFERENCE.md and TROUBLESHOOTING.md, then restart the host.

Read the command result carefully for manual-setup messages
Check /mcp in Claude Code or /plugin list for plugins
Re-run the matching connector after upgrading agentmemory

What people use it for

Connecting agentmemory to Claude Code so the agent can save and recall project context
Wiring agentmemory into Cursor for MCP tools, then adding its plugin for lifecycle capture
Setting up Pi or Oh My Pi with a running daemon and native recall
Checking which of the 22 supported agents agentmemory can connect to
Fixing a connector that wrote its config to the wrong path

Frequently asked questions

How do I install the agentmemory agent skill?

Download the ZIP from this page and unzip it into ~/.claude/skills/ for personal use, or into a project's .claude/skills/ folder. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.

Which agents does agentmemory support?

The skill's REFERENCE.md lists 22 connect adapters with names such as claude-code, cursor, codex and gemini-cli. Check that file for the full list and protocol notes for each one.

Do I need the agentmemory CLI installed first?

Yes. The skill requires the agentmemory CLI to be installed and available on the command line before you run agentmemory connect.

Does the connect command change my config files?

Yes. It merges the memory server into the host agent's configuration and keeps existing servers, but it does edit those files. Review the changes before relying on them.

Why don't the agentmemory tools show up after connecting?

Restart the host or run its MCP reload, such as /mcp in Claude Code, then check the tool list. Native memory and MCP connections are separate, so confirm the MCP tools are actually listed.

Can cloud-hosted agents use agentmemory?

Not with a local server. The skill notes that cloud Devin sessions cannot reach a local agentmemory server, so the plugin is for local sessions only.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits developers who run agentmemory locally and want their coding agent connected to it without hand-editing config files. It is not useful if you do not use agentmemory or a supported host, or if you cannot run a local server. Review the config changes the connect command makes before you rely on them.

Download the skill