agentmemory-config skill for Claude Code: set up and diagnose agentmemory
Helps Claude check and fix a local agentmemory setup, covering the MCP link, ports, environment variables and feature flags, for agentmemory users.
- Installs
- 11.2K
- on skills.sh
- Stars
- 30.1K
- on GitHub
- Files
- 2
- besides SKILL.md
- Updated
- Oct 2026
- last repo commit
agentmemory-config is a reference skill that Claude loads when you ask about getting started, a broken connection, enabling a feature or configuring auth. It walks through a fixed check order: read the agentmemory://status resource (or call memory_sessions), check Node.js 20+ and agentmemory status, then sort out which MCP connection and which secret is in use. It also lists the opt-in features and default ports.
The SKILL.md sets user-invocable: false, so it is not a slash command. It bundles no scripts, and it needs the agentmemory server running. It tells Claude not to save a test memory, change global host configuration, enable paid features or import transcripts unless you authorized it.
Good for
- People setting up agentmemory for the first time
- Anyone whose memory tools are missing or not connecting
- Users who want to know why a feature such as compression is off
Install agentmemory-config
Claude Code
- 1
Copy this prompt.
Install the agentmemory-config skill in Claude Code. Run this command: npx -y @agentmemory/agentmemory@latest Then check that it loaded and tell me how to use it.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires Node.js 20 or newer.
This skill helps with an agentmemory setup, so it needs the memory server and the plugin. The plugin route is the one the README recommends for Claude Code. The skills-only route installs all 17 skills and does not connect the MCP tools by itself. Pick one route, not both.
From the project README. Source.
What's in the skill
When it loads
Set up and diagnose local agentmemory, including MCP, environment variables, ports, and feature flags. Use when getting started, checking a connection, enabling a feature, configuring auth, or explaining why a feature is off by default.
Sections in SKILL.md
- Check the connection first
- Optional richer memory
- Defaults worth knowing
- Ports
- See also
- Reference
Key facts
- Author
- Rohit Ghumare (rohitg00)
- Repository
- rohitg00/agentmemory
- Installs on skills.sh
- 11,192 as of 2026-11-02
- GitHub stars
- 30,069 for the whole rohitg00/agentmemory repo
- License
- Apache-2.0 (plugin.json)
- Last repo commit
- 2026-10-29
What it includes
Six ordered checks of MCP, daemon, secret, mode and hooks
Claude reads agentmemory://status or calls memory_sessions, runs agentmemory status, and checks which secret and host connection are in use. It also reads /hooks before it promises automatic capture.
Try: “Check that agentmemory is connected.”
Runs code ·
Check the connection firstTurn on AGENTMEMORY_AUTO_COMPRESS and AGENTMEMORY_INJECT_CONTEXT in ~/.agentmemory/.env
This happens after you choose a provider. SKILL.md says to set its key locally without printing it.
Try: “Turn on automatic compression for agentmemory.”
Write ·
Optional richer memoryKeyword recall needs no key; compression and hook injection are opt-in
Local embeddings need explicit setup and model downloads. AGENTMEMORY_TOOLS is all by default or core.
Try: “Why is context injection off in agentmemory?”
Defaults worth knowingREST on 3111, viewer on 3113, engine WebSocket on 49134
Use --port or --instance to choose another local runtime. agentmemory status shows health and the viewer URL.
PortsList of 75 recognized AGENTMEMORY_* variables, and the state backend
The state backend is file by default or redis, which needs a plain redis:// URL.
Read ·
REFERENCE.mdRecovery steps when MCP tools are missing, and a REST fallback
The REST fallback calls the daemon with a bearer secret. It maps each skill to a REST endpoint.
Runs code ·
TROUBLESHOOTING.md
From the official docs
Permissions
Medium risk
The skill reads your local agentmemory secret and can edit ~/.agentmemory/.env, in a plugin that also records your sessions through hooks. Approve each config change and keep provider keys out of chat.
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
It reads your agentmemory secret files
The troubleshooting steps tell the assistant to load the secret from AGENTMEMORY_SECRET, then ~/.agentmemory/.env, then ~/.agentmemory/secret, without displaying it. Set AGENTMEMORY_CREDENTIAL_MODE=explicit to turn off local credential discovery. github.com
Enabling features sends content to your LLM provider
SKILL.md says enabled external model providers can receive selected content. Read that boundary before you set a provider key or turn on compression. github.com
The plugin records your sessions through hooks
The README lists captured data: user prompts (privacy-filtered), tool input and output, and the assistant reply of each turn. Review what is stored in the viewer, and use the forget skill to remove entries. github.com
Two connections can double-capture
SKILL.md warns that duplicate native and global hooks can double capture, and says to let you pick one MCP connection if both are configured. github.com
Technical details
| Turning on read-only | No read-only switch is documented for this skill. |
|---|---|
| Talks to | The agentmemory server on localhost:3111 (REST) and localhost:3113 (viewer). An LLM provider you configure, which can receive selected content once compression or consolidation is enabled |
| Sign-in and keys | A local server makes a random secret at ~/.agentmemory/secret. A remote server needs AGENTMEMORY_SECRET in the host environment, and the skill says local secrets are never forwarded to remote URLs. |
| Scopes |
|
Questions
What does the agentmemory-config skill do?
It guides Claude through checking and configuring a local agentmemory. That covers the MCP connection, environment variables, ports, auth and which features are on or off.
How do I install agentmemory-config?
Use the install box at the top of this page. Start the agentmemory server first, then add the plugin. Installing the skill folder alone is not enough.
How do I use it?
Ask for the task, such as "check that agentmemory is connected". The SKILL.md sets user-invocable to false, so it is not a slash command. Claude loads it when your request matches its description.
Does it need an API key?
Not for basic keyword recall, which needs no key. LLM compression and semantic search need a provider key you set yourself in ~/.agentmemory/.env.
Why is a feature off by default?
Per-observation compression (AGENTMEMORY_AUTO_COMPRESS) and hook context injection (AGENTMEMORY_INJECT_CONTEXT) are opt-in. SKILL.md says enabled providers can receive selected content, so you turn them on yourself.
Which ports does agentmemory use?
REST uses 3111 and the viewer uses 3113. The internal engine WebSocket uses 49134. Use --port or --instance to pick another local runtime.
Is it safe to use?
It reads local secret files and can edit ~/.agentmemory/.env, and the plugin around it records your sessions through hooks. SKILL.md says Claude should not change global host config, enable paid features or import transcripts without your approval. Approve changes one at a time.
Information
updated Nov 2, 2026
- Repository
- rohitg00/agentmemory
- Folder
plugin/skills/agentmemory-config
Sources
Docs, issue trackers and discussions this page is built from.
Show all 7Hide
- agentmemory README · github.com
- agentmemory-config SKILL.md · github.com
- agentmemory-config REFERENCE.md · github.com
- agentmemory plugin.json · github.com
- agentmemory-config TROUBLESHOOTING.md · github.com
- agentmemory plugin .mcp.json · github.com
- agentmemory plugin hooks.json · github.com