By wandb
Track Claude Code sessions in Weave for observability and debugging. Automatically captures all Claude Code events via hooks, and provides settings management for project, API key, and agent name. Supports installation, upgrade, and status diagnostics.
This skill should be used when the user wants to "configure weave", "set weave project", "change weave project", "set wandb api key", "update weave settings", "show weave config", "change weave configuration", "restart the weave daemon", "apply weave config changes", "restart weave to pick up changes", or needs to read or update any Weave Claude Code plugin settings.
This skill should be used when the user wants to "install the weave plugin", "set up weave", "install weave-claude-code", "configure weave for the first time", "get started with weave tracing", or needs to complete the initial setup of the Weave Claude Code plugin including dependency installation and project configuration.
This skill should be used when the user wants to "check weave status", "verify the weave plugin is running", "see if weave is set up correctly", "check weave configuration", "is weave working", "weave is running an older config", "the daemon is on an old config", or needs to diagnose why Claude Code sessions are not appearing in Weave.
Matches all tools
Hooks run on every tool call, not just specific ones
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Track Claude Code sessions in Weave for observability and debugging. Every session, turn, tool call, and subagent is automatically logged as a structured trace — no code changes required.
1. Install the CLI
npm install -g weave-claude-code
2. Run the installer
weave-claude-code install
This will:
~/.weave-claude-code/settings.jsonentity/project) and W&B API key if not already setYour W&B API key is available at https://wandb.ai/authorize.
For CI, bootstrap scripts, or other automated systems, you can skip prompts:
WEAVE_PROJECT=my-entity/my-project \
WANDB_API_KEY=<your-api-key> \
weave-claude-code install --non-interactive
In non-interactive mode, the installer still creates config, registers the Claude marketplace, and installs the plugin. It does not prompt for missing values. Instead, it:
WEAVE_PROJECT and WANDB_API_KEY from the environment when presentsettings.jsonBy default, claude plugin marketplace add clones wandb/weave-claude-code from GitHub. In CI or container sandboxes without git/SSH access to GitHub, pass --source=local to register the marketplace from the npm-installed tree on disk instead:
npm install -g weave-claude-code
weave-claude-code install --non-interactive --source=local
--source=local requires the package to be installed globally via npm first (it reads from $(npm root -g)/weave-claude-code). Upgrades follow the npm cadence; the marketplace ref drift check is skipped for local sources.
3. Restart or launch Claude Code
If Claude Code is already running, reload plugins from inside the session:
/reload-plugins
Otherwise, launch Claude Code from any folder:
claude
Sessions are traced automatically from this point, across all Claude Code sessions on this machine. Tracing is not scoped to a single project. To stop tracing, run weave-claude-code uninstall.
Open your Weave project to see them.
This plugin sends Claude Code session data to W&B Weave.
That data can include sensitive content, including:
If Claude Code accesses secrets, credentials, proprietary source code, personal data, or other confidential material during a session, that information may be logged to W&B Weave as part of the trace.
PII scrubbing and sensitive-data redaction are not yet implemented in the current version. If you cannot send this data to W&B Weave under your security or compliance requirements, do not install or enable this plugin yet.
# Show all current settings (env-var overrides are flagged in the output)
weave-claude-code config show
# Read a single setting (resolves env-var overrides)
weave-claude-code config get weave_project
# Set your Weave project
weave-claude-code config set weave_project my-entity/my-project
# Set your W&B API key
weave-claude-code config set wandb_api_key <your-api-key>
# (Optional) Customize the agent name shown in Weave's Agents view (default: claude-code)
weave-claude-code config set agent_name my-team-bot
You can also set these via environment variables — they take precedence over the settings file:
export WEAVE_PROJECT=my-entity/my-project
export WANDB_API_KEY=<your-api-key>
export WEAVE_AGENT_NAME=my-team-bot
This is especially useful with weave-claude-code install --non-interactive, where the installer checks these variables instead of prompting.
If you use W&B Dedicated Cloud or a self-hosted instance, set WANDB_BASE_URL to point the plugin at your deployment before launching Claude Code:
export WANDB_BASE_URL=https://your-instance.wandb.io
On SaaS, leave WANDB_BASE_URL unset (traces go to https://trace.wandb.ai). Setting it to the SaaS API host https://api.wandb.ai is handled automatically, since that host has no trace-ingest route. To send traces somewhere other than the wandb API host, set WF_TRACE_SERVER_URL, which takes precedence over WANDB_BASE_URL.
npx claudepluginhub wandb/weave-claude-code --plugin weaveUltra-compressed communication mode. Cuts 65% of output tokens (measured) while keeping full technical accuracy by speaking like a caveman.
Memory compression system for Claude Code - persist context across sessions
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
45% cost reduction measured. Cache expiry prevention, SubTask auto-delegation, zero-cost context restoration, real-time cost dashboard. The only Claude Code plugin built from CC source analysis.