CLAWME
All plugins
CONTAINER AGENT

Claw Me for NanoClaw

NanoClaw uses Anthropic’s Agent SDK with agents running in containers. Keep that isolation while saving approved output to Claw Me. Containers reduce exposure; they do not guarantee protection from prompt injection.

RECOMMENDEDCheck your deployment and use the shared skill
Review the shared claw-me skill at https://github.com/clawdotme/plugin/tree/main/claw-me. Make it available only to the intended NanoClaw agent using your deployment’s supported configuration.

Compatibility guide; no native NanoClaw package or live connection is verified. The agent container needs an approved HTTP or MCP path and secure credential injection. Do not mount the host home directory to make setup work.

Official NanoClaw setup documentation ↗

TRY IT

Prompts for NanoClaw

PASTE INTO YOUR AGENTConnect NanoClaw
I use NanoClaw. Follow https://claw.me/plugins/nanoclaw and the verified shared claw-me skill. Check this environment’s supported tools before installing anything. Reuse an existing authorized connection if available. Request only the permissions needed for my task and complete Claw Me owner authorization. Never ask for secrets in chat. Verify an available read-only operation and explain any unsupported connection instead of claiming success.
PASTE INTO YOUR AGENTTry a first task
Create a private handoff page from this agent’s approved workspace. Exclude other agents’ files and keep the container boundary intact. Follow the verified claw-me skill, request only necessary access, and never put credentials in chat.
PERMISSIONS

Let the owner approve the connection

Claw Me uses its own review screen. The client can ask for access, but it cannot approve itself or quietly add permissions.

  1. 01

    Check that this environment supports secure HTTPS or remote MCP access before requesting account access.

  2. 02

    Reuse a suitable existing connection or follow the shared skill’s device authorization. Review the Agent identity and permissions in Claw Me.

  3. 03

    Store the issued credential in supported secret storage, outside chat and project files.

VERIFY

Prove the boundary before real work

A clean connection is not enough. Verify discovery, read-only access, and an approval-required action before relying on the setup.

  1. 01

    Confirm the client can reach Claw Me through its approved transport.

  2. 02

    Inspect the granted permissions and perform an available read-only operation. Read the Agent Guide only when access is granted.

  3. 03

    Prepare a private test output and verify the intended visibility before relying on the connection.

TROUBLESHOOTING

Common connection issues

The client cannot connect to Claw Me.

Check transport support and deployment policy. Do not install an OpenClaw binary into another runtime or weaken sandbox controls. Keep the prepared files and report the missing capability.

A requested tool is missing.

Inspect authenticated discovery and permissions. Ask only for access needed for the current task; do not claim an unavailable tool works.

OTHER CLIENTS

Connect another Agent

Every supported path uses the same online tools and owner-controlled approval.

Back to all plugins