Load the 1Claw plugin into an elizaOS agent

The @1claw/plugin-elizaos package gives an elizaOS character HSM-backed vault access and multi-chain signing. It wraps @1claw/sdk and exposes actions like GET_SECRET, LIST_SECRETS, and SIGN_MESSAGE.

1. 1

   Clone the plugin repo and install.

   bash Copy

   git clone https://github.com/1clawAI/1claw-elizaos-plugin.git
   cd 1claw-elizaos-plugin
   npm install

2. 2

   Run bootstrap with your human key to create a vault, agent, and policy.

   bash Copy

   ONECLAW_HUMAN_API_KEY=1ck_your_key npm run bootstrap

3. 3

   Bootstrap writes .env.elizaos with agent credentials. Validate them.

   bash Copy

   export $(grep -v '^#' .env.elizaos | xargs)
   npm run validate

4. 4

   Add the plugin to your character file. Copy ONECLAW_AGENT_API_KEY, ONECLAW_AGENT_ID, and ONECLAW_VAULT_ID from .env.elizaos into settings.secrets in your character file.

   json Copy

   {
     "name": "my-agent",
     "plugins": ["@1claw/plugin-elizaos"],
     "settings": {
       "secrets": {
         "ONECLAW_AGENT_API_KEY": "ocv_from_env_elizaos",
         "ONECLAW_AGENT_ID": "uuid-from-env-elizaos",
         "ONECLAW_VAULT_ID": "uuid-from-env-elizaos"
       }
     }
   }

5. 5

   Alternative: add to an existing project. If you already have an elizaOS project, install the plugin directly instead of cloning the repo.

   bash Copy

   npm install @1claw/plugin-elizaos

6. 6

   Start your elizaOS agent and verify the 1claw plugin loaded. You should see a log line confirming the plugin registered its actions.

   bash Copy

   npx elizaos start --character your-character.json

Your elizaOS agent now boots with the 1Claw plugin loaded, authenticates with its ocv_ key, and can run vault and signing actions.

What this gives you, and what it does not. The plugin registers actions an agent can take; the guardrails on those actions still live in 1Claw.

• Registering an action does not authorise it. The agent's policies and guardrails still apply, so an action can be present and consistently refused.
• Multi-chain signing is included. That raises the stakes: an elizaOS agent with signing actions needs the same transaction guardrails as any other signing agent.
• The context provider shapes what the model sees. What it injects becomes part of the prompt, so it is worth reading rather than assuming.