weave-openclaw

weave-openclaw

OpenClaw plugin for tracing agent runs, model calls, tool calls, tokens, and costs in W&B Weave.

[!WARNING] captureContent defaults to true. Prompts, replies, and tool inputs and results are sent unredacted to W&B. Set it to false to record only trace structure, tokens, and costs.

Requirements

• Node.js >= 22.14.0
• OpenClaw >= 2026.4.25
• A W&B account and project

Setup

Install the plugin:

openclaw plugins install clawhub:@wandb/weave-openclaw

Export a W&B API key:

export WANDB_API_KEY=<your-key>

Add the plugin to ~/.openclaw/openclaw.json:

{
  diagnostics: { enabled: true },
  plugins: {
    allow: ["weave"],
    entries: {
      weave: {
        enabled: true,
        config: { entity: "your-team", project: "your-project" },
        hooks: { allowConversationAccess: true },
      },
    },
  },
}

Restart the gateway if needed:

openclaw gateway restart

Run /weave status in a chat. When it reports running, view traces at:

https://wandb.ai/<entity>/<project>/weave/agents

hooks.allowConversationAccess: true allows prompts, replies, and per-call token counts. Without it, trace structure, tool calls, and run totals still work. diagnostics.enabled: false disables tracing.

See the full setup guide and ClawHub listing.

Configuration

Only entity and project are required. Everything else has a sensible default.

{
  plugins: {
    entries: {
      weave: {
        enabled: true,
        config: {
          entity: "your-team",        // your W&B team or username
          project: "your-project",    // your W&B project name

          // Leave apiKey out to use the WANDB_API_KEY environment variable.
          // File and exec SecretRefs also work with a configured OpenClaw
          // secret provider.
          // A plain key string works too, but keeping secrets out of config is safer:
          //   apiKey: "your-wandb-api-key"
          apiKey: { source: "env", provider: "default", id: "WANDB_API_KEY" },

          serviceName: "openclaw-agent",   // shown in /weave status
          // These help group and label your agents in the dashboard.
          agentName: "my-agent",
          agentVersion: "v1.0",
          agentDescription: "What my agent does.",

          // On by default. Set to false to stop recording the actual message
          // text (for example, to meet a privacy or retention policy). The
          // plugin records text as-is and does not hide sensitive values, so
          // remove them beforehand if you need to.
          captureContent: true,

          // How often (in milliseconds) traces are sent.
          flushIntervalMs: 1000,
        },
        hooks: { allowConversationAccess: true },
      },
    },
  },
}

Environment refs work without extra setup. File and exec refs need a matching secrets.providers entry; see OpenClaw secrets management.

Credential lookup order:

• apiKey SecretRef
• Plain apiKey
• WANDB_API_KEY
• ~/.netrc

OpenClaw also loads WANDB_API_KEY from ~/.openclaw/.env. Set WANDB_BASE_URL for dedicated or self-hosted W&B.

Troubleshooting

Problem	Check
/weave status is not running	Check entity, project, the plugin version, and gateway logs.
No traces	Ensure diagnostics are enabled and the configured project matches /weave status.
Blank messages or model calls	Set hooks.allowConversationAccess: true and restart.
401 or 403	Refresh the API key and confirm project access.
404 on self-hosted W&B	Check WANDB_BASE_URL.
Connection or DNS error	Check gateway network, proxy, and firewall access to W&B.

Manage the plugin

openclaw plugins update weave
openclaw plugins disable weave
openclaw plugins enable weave
openclaw plugins uninstall weave

Development

pnpm install --frozen-lockfile
pnpm check

openclaw plugins install --link .
openclaw gateway restart
openclaw plugins inspect weave --runtime --json

License

MIT