Weco DocsUse Claude Code in the dashboard
Bridge a local Claude Code session to the Weco dashboard with `weco start claude`. Drive your agent and watch your runs from the same place.
weco start claude connects your local Claude Code session to the Weco dashboard. Your agent runs natively in your terminal and its conversation streams live into the dashboard, where you can read along, message it back, and watch your runs in one place.
Launch Claude Code bridged to the dashboard
From your project directory:
weco start claudeThis starts Claude Code under a thin Weco wrapper and prints a dashboard URL. Open it and you'll see the session mirrored in the chat panel next to your run.
Drive from either side
Type in your terminal, or type in the dashboard chat. When you spot a plateau on the live chart, message the agent right there to steer the run:
Branch from the best node and add engineered interaction features for the top-importance columns, 12 steps.The agent acts locally, and the result appears in the run's tree as it explores.
Stay in the loop while it runs
Long-running evals don't freeze the conversation. The bridge tracks the run's progress and feeds status changes back to the agent, so it keeps you posted, and you can interject at any time without losing the session.
Useful flags
| Flag | What it does |
|---|---|
--headless | Run with no local TUI, streaming only to the dashboard. Pair with --prompt to seed the first turn and --allow-tools for unattended runs. |
-p, --prompt | Seed the first turn with a prompt instead of waiting for input. |
--allow-tools | Auto-approve the agent's tool calls (no per-action prompts). Use only when you trust the session. |
--billing weco | Route the agent's LLM calls through Weco's proxy and bill your Weco credit wallet instead of your local Claude auth. |
For the full flag list, see the CLI Reference.
weco start claude requires the Claude Code CLI to be installed, since it runs Claude Code under the hood.
See also
- Steerability — branch and steer runs while you watch
- Watch, share & manage runs — the dashboard surface
Steerability: Branching and Subtrees
Steer a Weco autoresearch run mid-flight. When a run plateaus or you have an idea to try, branch from where it already is instead of starting over. Each steer becomes a new subtree in the search.
Weco Observe: Track any Optimization Loop
Use Weco Observe to track experiments from your own optimization loop — an LLM agent, a training script, or manual runs — with tree visualization, code diffs, and metrics.