Guide · Agent skill
How to use the Add Mnemon Persistent Memory agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The add-mnemon skill gives agents running in NanoClaw containers a persistent memory layer built on mnemon. Before answering, the agent can recall relevant past context, and after each turn it can store new insights. This guide explains what the skill changes, how to install it, how to check that it works, and where its risks are.
Add Mnemon Persistent Memorynanocoai/nanoclaw · MIT Download the skill Quick start
- 1Check the provider. Run the grep on groups/*/container.json and confirm each group you want to change uses the default Claude provider.
- 2Install the skill folder. Download the ZIP from this page and unzip it into ~/.claude/skills/ or your project's .claude/skills/, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 3Apply the changes. Ask your agent to apply add-mnemon in your NanoClaw project. It adds the Dockerfile install block, the mnemon setup line in the entrypoint, and the two integration tests.
- 4Rebuild the image. Run ./container/build.sh and confirm mnemon --version runs inside the nanoclaw-agent image.
- 5Restart and verify. Restart the NanoClaw service, check container logs for mnemon, and test recall by starting a new session that references an earlier conversation.
What is an agent skill and what does add-mnemon do?
An agent skill is a folder containing a SKILL.md file with instructions that an AI coding agent follows to make a specific change. The add-mnemon skill's SKILL.md describes how to add the mnemon memory tool to a NanoClaw project.
Once applied, mnemon is installed in the agent container image. On each container start, mnemon setup registers Claude Code hooks that surface relevant memory before the agent responds and store new insights after each turn. Memory is written to the per-agent-group .claude/ mount, so it survives container restarts.
- Installs the mnemon binary into the agent container image
- Runs mnemon setup on each container start via the entrypoint
- Registers hooks in /home/node/.claude/settings.json
- Stores memory in the agent group's .claude folder on the host
Check that your agent group uses the Claude provider
mnemon hooks only fire under the Claude Code target. The skill is for agent groups on the default Claude provider. Groups that set a different provider, such as opencode, spawn their own process and never invoke the claude CLI, so the hooks will not run for them.
Check each group's container.json before applying. No match, or a provider value of claude, means the group uses Claude.
grep -H '"provider"' groups/*/container.json 2>/dev/nullInstall the add-mnemon skill
Download the add-mnemon ZIP from this page. To make it available to Claude Code, unzip it into ~/.claude/skills/ for all projects, or into a project's .claude/skills/ folder for one project. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
Then ask your agent to apply the skill in your NanoClaw project root. The source does not define fixed trigger phrases, so a plain request such as asking the agent to apply the add-mnemon skill is the practical approach. The skill's steps are written for a NanoClaw project that contains container/Dockerfile and container/entrypoint.sh.
- Dockerfile: insert the mnemon install block above the '# ---- Bun runtime' section of container/Dockerfile
- Entrypoint: add mnemon setup --target claude-code --yes --global right after set -e in container/entrypoint.sh
- Tests: copy mnemon-dockerfile.test.ts and mnemon-entrypoint.test.ts into src/ and run them with pnpm exec vitest
cp .claude/skills/add-mnemon/mnemon-dockerfile.test.ts src/mnemon-dockerfile.test.ts
cp .claude/skills/add-mnemon/mnemon-entrypoint.test.ts src/mnemon-entrypoint.test.ts
pnpm exec vitest run src/mnemon-dockerfile.test.ts src/mnemon-entrypoint.test.tsRebuild the image and restart the service
Rebuild the agent image so the Dockerfile layer is included, then smoke-test that the mnemon binary runs. Then restart NanoClaw from the project root. Use the systemd line on Linux or the launchd line on macOS.
The skill notes that every step is idempotent, so re-running it on an already applied project is safe.
./container/build.sh
docker run --rm --entrypoint mnemon nanoclaw-agent:latest --version
source setup/lib/install-slug.sh
systemctl --user restart $(systemd_unit)
# launchctl kickstart -k gui/$(id -u)/$(launchd_label) # macOSConfirm recall works and inspect the hooks
After the next container starts, check the logs for mnemon output and inspect the hooks inside the running container. Then test recall: have a conversation with the agent, start a new session, and refer to something from the earlier one. Mnemon should surface the relevant context without you restating it.
To find where memory lives on the host, use docker inspect on the running container. To reset an agent's memory, stop the container and delete the mnemon/ subdirectory at that host path.
- Memory path inside the container: /home/node/.claude/mnemon/
- Host path: the source of the /home/node/.claude mount
docker logs $(docker ps --filter label=nanoclaw-session --format "{{.Names}}" | head -1) 2>&1 | grep -i mnemon
docker exec $(docker ps --filter label=nanoclaw-session --format "{{.Names}}" | head -1) cat /home/node/.claude/settings.json | grep -A5 mnemonTroubleshoot common problems
If mnemon: command not found appears in the container, the image was not rebuilt after the Dockerfile change. If memory does not persist, check that MNEMON_DATA_DIR resolves to a mounted path. If the agent ignores past memory, check whether the hooks exist in settings.json. If they are missing, mnemon setup may have failed, so check the container startup logs. To see the full error, run setup manually inside a running container.
- Command not found: rebuild with ./container/build.sh and restart
- Memory not persisting: confirm the mount, not an ephemeral directory
- Hooks missing: check container startup logs for mnemon errors
- Setup failure: run mnemon setup manually inside the container
docker exec -it <container> mnemon setup --target claude-code --yes --globalWhat people use it for
- Keeping project context across sessions so you do not restate decisions each time
- Letting a support-style agent group recall earlier customer details from its own memory folder
- Reviewing or resetting what a developer's agent group has remembered by inspecting or deleting the mnemon folder on the host
Frequently asked questions
How do I install the add-mnemon agent skill?
Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps. Then ask your agent to apply it in a NanoClaw project that has container/Dockerfile and container/entrypoint.sh.
What do I need to run the add-mnemon skill?
You need a NanoClaw project with container/Dockerfile and container/entrypoint.sh, Docker, and access to GitHub to download the mnemon release binary and query the GitHub API. The included tests need pnpm and vitest.
Is the add-mnemon skill free and what is its licence?
The item is listed under the MIT licence. The source does not state pricing beyond that, so check the repository for the terms that apply to your use.
Does add-mnemon work with opencode or other providers?
No. mnemon hooks only fire under the Claude Code target, so agent groups using a non-Claude provider such as opencode will never run them. Use the skill only on groups that run the default Claude provider.
Why does mnemon: command not found appear?
The container image was probably not rebuilt after the Dockerfile layer was added. Run ./container/build.sh, then restart the service.
How do I reset what the agent remembers?
Stop the agent container, find the host path of the /home/node/.claude mount with docker inspect, and delete the mnemon/ subdirectory inside it.
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
Use add-mnemon if you run NanoClaw agent groups on the default Claude provider and want conversations to carry over, and you are comfortable reviewing a folder on disk. Pick something else if your groups use a non-Claude provider, or if you handle sensitive information and have not reviewed what gets stored, since memory persists and is surfaced back into future prompts.
Download the skill