@1claw/plugin-elizaos

HSM-backed secrets and multi-chain signing for elizaOS agents via 1Claw.

Your agent gets vault access and signing keys without ever seeing private keys or holding credentials in context.

Install

npm install @1claw/plugin-elizaos
# or
bun add @1claw/plugin-elizaos

Quick Start

New to 1Claw? Run bootstrap first to create a vault, agent, and policy with a human 1ck_ key, then use the generated ocv_ key below.

Add to your character's plugins:

{
  "name": "my-agent",
  "plugins": ["@1claw/plugin-elizaos"],
  "settings": {
    "secrets": {
      "ONECLAW_AGENT_API_KEY": "ocv_your_key_here"
    }
  }
}

The plugin authenticates on boot, discovers the agent's vault, and injects secret paths and signing keys into context each turn.

Bootstrap (first-time setup)

Use a human API key (1ck_... from 1claw.xyz/dashboard → API Keys) only to create a vault, agent, and access policy. The script never writes the human key to disk — only the one-time agent key (ocv_...) is saved.

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

# Provision vault + agent + policy (human key used once, in memory only)
ONECLAW_HUMAN_API_KEY=1ck_your_key_here npm run bootstrap

# Optional: enable Intents API signing at create time
ONECLAW_HUMAN_API_KEY=1ck_... ONECLAW_ENABLE_INTENTS=true npm run bootstrap

Outputs:

• .env.elizaos — agent-only env vars (ONECLAW_AGENT_API_KEY, ONECLAW_AGENT_ID, ONECLAW_VAULT_ID)
• Terminal — one-time ocv_... key and a character JSON snippet

Environment variables (bootstrap):

| Variable | Default | Description | |---|---|---| | ONECLAW_HUMAN_API_KEY | (prompt) | Human key 1ck_... — setup only, not stored | | ONECLAW_AGENT_NAME | elizaos-agent | Agent display name | | ONECLAW_VAULT_NAME | elizaos-vault | Vault name | | ONECLAW_POLICY_PATH | ** | Glob for granted secret paths | | ONECLAW_ENABLE_INTENTS | false | Set true to enable transaction signing | | ONECLAW_OUTPUT_FILE | .env.elizaos | Where to write agent credentials | | ONECLAW_BASE_URL | https://api.1claw.xyz | API base URL |

Validate agent credentials after bootstrap:

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

Configuration

| Variable | Required | Default | Description | |---|---|---|---| | ONECLAW_AGENT_API_KEY | Yes | — | Agent API key from 1claw.xyz/dashboard, format ocv_... | | ONECLAW_AGENT_ID | No | auto | Override auto-discovery (rarely needed) | | ONECLAW_VAULT_ID | No | auto | Pin a specific vault when agent has multiple | | ONECLAW_BASE_URL | No | https://api.1claw.xyz | API endpoint | | ONECLAW_USE_SHROUD | No | false | Route through Shroud TEE proxy at shroud.1claw.xyz |

Actions

| Action | Description | |---|---| | GET_SECRET | Fetch a secret by path. Value goes to tool channel, never chat. | | LIST_SECRETS | List available paths (no values). | | PUT_SECRET | Store or update a secret. | | SIGN_MESSAGE | EIP-191 personal sign (agent's signing key never leaves HSM). | | SIGN_TYPED_DATA | EIP-712 structured data signing. | | SIMULATE_TRANSACTION | Tenderly dry-run before broadcasting. | | SUBMIT_TRANSACTION | Sign + broadcast EVM transactions with server-side guardrails. | | LIST_SIGNING_KEYS | Show provisioned chains and public addresses. | | EXECUTE_HTTP | Execute an HTTP request through a pre-configured binding (Pro+, requires execution_intents_enabled). | | LIST_BINDINGS | List available bindings (credential handles for external services). |

Security Model

• The agent only sees what the human's policies grant. An ocv_ API key has zero access by default — a human must create policies on the 1Claw dashboard.
• Secret values retrieved via GET_SECRET are returned in the action's data payload but marked [REDACTED] in the user-facing text.
• Private keys for signing never leave the HSM/TEE. The agent submits intents; 1Claw signs server-side.
• Per-agent guardrails (address allowlists, value caps, daily limits, chain restrictions) are enforced server-side before any signing occurs.

Example Character

{
  "name": "DeFi Agent",
  "plugins": ["@1claw/plugin-elizaos"],
  "bio": "I manage on-chain positions using 1Claw for secure signing.",
  "settings": {
    "secrets": {
      "ONECLAW_AGENT_API_KEY": "ocv_abcdef123456"
    }
  }
}

Development

git clone https://github.com/1clawAI/1claw-elizaos-plugin.git
cd 1claw-elizaos-plugin
npm install
npm run build
npm test
chmod +x scripts/*.sh

# Live bootstrap test (needs ONECLAW_HUMAN_API_KEY or ADMIN_EMAIL+ADMIN_PASSWORD in repo root .env)
npm run test:integration

Execution Intents (v0.2)

Execution Intents let agents make HTTP calls to external services through pre-configured bindings. Credentials are stored server-side and never exposed to the agent.

Requirements: execution_intents_enabled must be toggled on for the agent, and the org needs a Pro+ subscription.

{
  "execution": {
    "binding": "github-api",
    "method": "GET",
    "path": "/user/repos"
  }
}

Bindings are created by humans via the dashboard or SDK — agents can only list and use them.

Roadmap

• Treasury wallet actions (delegated-treasury flow)
• Platform API actions (multi-tenant provisioning)
• Shroud as model-provider plugin (plugin-1claw-shroud)
• CMEK / MPC client-share handling

Links

• 1Claw Platform
• Documentation — elizaOS guide
• Integration Guide
• Ecosystem
• Shroud TEE Proxy
• Intents API
• @1claw/plugin-elizaos on npm
• @1claw/sdk on npm

License

MIT