Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.wandb.ai/llms.txt

Use this file to discover all available pages before exploring further.

Weave for Agents is in public preview. Features, APIs, and the Agents view UI may change before general availability.
The Weave OpenClaw plugin automatically traces every agent session running through the OpenClaw gateway and sends the structured data to W&B Weave. Every turn, model call, and tool execution is logged with no application code changes required. Use these traces to debug sessions, audit tool usage, and monitor cost and latency across runs.
This plugin sends OpenClaw session data to Weave. That data can include user prompts, model responses, tool inputs and outputs, tool results, and conversation history.PII scrubbing and sensitive-data redaction aren’t implemented by the plugin. If you need to suppress content capture, set captureContent: false in the plugin config. If you can’t send this data to Weave under your security or compliance requirements, don’t install this plugin.

​
Prerequisites

  • Node.js v22 or later.
  • OpenClaw >= 2026.4.25 with the plugin API.
  • A W&B account and API key.
  • A Weave project ([YOUR-TEAM]/[YOUR-PROJECT]) to receive traces.

​
Install the plugin

1

Install the package:

pnpm add weave-openclaw
The plugin is loaded by the OpenClaw gateway through its config. You don’t import it from application code.
2

Add the plugin to your gateway config:

The default config location is ~/.openclaw/openclaw.json (JSON5, so comments and trailing commas are allowed). Run openclaw onboard to scaffold one if you don’t have it yet. Update [YOUR-TEAM] and [YOUR-PROJECT] for your project.
{
  diagnostics: { enabled: true },
  plugins: {
    allow: ["weave"],
    entries: {
      weave: {
        enabled: true,
        config: { entity: "YOUR-TEAM", project: "YOUR-PROJECT" },
        hooks: { allowConversationAccess: true },
      },
    },
  },
}
As shown in the example config, your gateway config must include these two settings:
  • Set diagnostics.enabled to true so OpenClaw emits the diagnostic events the plugin consumes.
  • Set hooks.allowConversationAccess to true so OpenClaw runs the content-bearing hooks (llm_input, llm_output, agent_end) and spans include input/output text and tool arguments and results.
3

Restart the gateway and verify:

Restart the OpenClaw gateway, then run /weave status in any OpenClaw chat surface to confirm the plugin is active. Traces will appear at wandb.ai/[YOUR-TEAM]/[YOUR-PROJECT] within a few seconds of your first agent run.

​
View OpenClaw traces in Weave

After running at least one agent session, open your project in the Weave UI:
  1. Navigate to https://wandb.ai and select your project.
  2. In the sidebar, select Agents for the multi-turn chat view and per-agent version grouping, or select Traces for the raw span tree.
  3. Select a conversation to inspect the full session tree.
For more on the Agents view, see View agent activity. The plugin emits spans per the OTel GenAI semantic conventions:
SpanEmitted onKey attributes
invoke_agent <agent>Per agent rungen_ai.agent.name, gen_ai.conversation.id, cumulative cost, token usage
chat <model>Per model callgen_ai.request.model, gen_ai.usage.input_tokens, gen_ai.usage.output_tokens
execute_tool <tool>Per tool executiongen_ai.tool.name, gen_ai.tool.call.id

​
Configuration reference

This section is the full configuration reference for the weave plugin entry in openclaw.json. The apiKey field supports four auth sources, resolved in this order:
  1. A SecretRef object with source: "env" or source: "file" (see line 10 in the following example).
  2. A literal apiKey string (supported but not recommended).
  3. The WANDB_API_KEY environment variable.
  4. A ~/.netrc entry for the Weave host, populated by wandb login.
{
  plugins: {
    entries: {
      weave: {
        enabled: true,
        config: {
          entity: "YOUR-TEAM",
          project: "YOUR-PROJECT",
          // Reads WANDB_API_KEY from env if apiKey is omitted.
          // SecretRef supports source: "env" or "file":
          //   { source: "env",  provider: "default", id: "WANDB_API_KEY" }
          //   { source: "file", provider: "default", id: "/run/secrets/wandb" }
          // Plain string is supported but discouraged.
          apiKey: { source: "env", provider: "default", id: "WANDB_API_KEY" },
          serviceName: "openclaw-agent",
          // Optional, improves Agents tab grouping.
          agentName: "my-agent",
          agentVersion: "v1.0",
          agentDescription: "What my agent does.",
          // ON by default. Set to false for a hard off (compliance or
          // retention policy). The plugin does not redact captured
          // strings; scrub upstream if needed.
          captureContent: true,
          flushIntervalMs: 5000,
        },
        hooks: { allowConversationAccess: true },
      },
    },
  },
}
captureContent is true by default. When captureContent is true, input and output messages, tool arguments, and tool results are also emitted following the gen_ai.input.messages / gen_ai.output.messages payload shape. Subagents, compaction events, loop detection, retry attempts, and context sizing are stamped as additional attributes and span events. Set captureContent to false to turn off capture for compliance or retention policies.

​
W&B Dedicated Cloud or self-hosted instances

The plugin delegates endpoint and auth to the Weave Node SDK. It reads the following environment variables using the same conventions as the Weave Python and Node SDKs:
VariableDescription
WANDB_BASE_URLW&B API base URL. Default: https://api.wandb.ai. Set this for Dedicated Cloud or self-hosted installs.
WF_TRACE_SERVER_URLFull trace-server URL override. Use for self-managed or proxied setups.

​
Troubleshooting

The gateway log is the terminal output from the process running openclaw, or your process manager’s log stream if you daemonized it.

​
Plugin loaded but no spans show up

  1. Run /weave status. If lifecycle is disabled, config-error, or not-started, the plugin did not activate. Check the gateway log for weave: config.entity is required, weave: configuration error, or [weave] incompatible plugin SDK.
  2. Confirm diagnostics.enabled: true in the gateway config.
  3. Confirm the entity and project match the URL slug of the Weave project you’re inspecting. /weave status should print project=<entity>/<project>.
  4. Confirm the auth source. /weave status should print auth=.... If it says WANDB_API_KEY env but you set the key in a different environment variable, the plugin is reading the wrong key.

​
Spans land but input/output text is empty

Look in the gateway log for: 
[plugins] typed hook "llm_input"  blocked because non-bundled plugins must set
                                  plugins.entries.weave.hooks.allowConversationAccess=true
[plugins] typed hook "llm_output" blocked ...
[plugins] typed hook "agent_end"  blocked ...
OpenClaw gates content-bearing hooks behind an operator opt-in. Set plugins.entries.weave.hooks.allowConversationAccess: true in your config and restart the gateway. Span structure and cost/usage data come through diagnostic events, not hooks, so those keep working even without the gate flipped.

​
Errors sending traces to Weave

If the plugin is active and spans are being generated but not appearing in Weave, check the gateway log for an export error and match it against the table below.
SymptomMost likely causeFix
401 / 403 from trace.wandb.aiInvalid or scope-limited API keyVerify the key is current and the team owns the entity/project. wandb login refreshes ~/.netrc.
404 from the agents endpointWrong base or trace-server URLFor dedicated installs, set WANDB_BASE_URL to your install host. For self-managed or proxy, set WF_TRACE_SERVER_URL to the trace-server URL.
Connection refused / DNS errorDNS, proxy, or firewallConfirm the gateway host can reach trace.wandb.ai (cloud) or your install host (dedicated) on 443.