Add Claude Opus 4.7 support
TL;DR
Default Anthropic selections, Opus aliases, and CLI defaults to Claude Opus 4.7 for improved model performance.
## What changed OpenClaw added default support for Claude Opus 4.7 on 18 May 2026. Anthropic model selections now route to Opus 4.7 by default. CLI aliases and configuration files updated to point at the new model without extra flags.
The change improves output quality on reasoning tasks and long context work. Users who connect an Anthropic key see the switch automatically after the next update.
## Why it matters Better default models raise the floor for what a self-hosted agent can handle without constant prompt tweaks. Vibe Builders gain more reliable browser control and skill writing, yet they also face higher per-token costs from Opus usage.
The move pressures competing agents that still default to lighter models. It bets that users will accept the spend for fewer failed runs on complex jobs.
## How to use it Pull the latest OpenClaw release through the existing CLI command. Edit the YAML file to confirm anthropic.model equals opus-4-7 or leave it blank to accept the new default. Restart the agent and test a single scheduled task to verify the model switch.
Set a hard spend cap in the Anthropic console before running heavier sessions. Monitor the first week of token usage through the provider dashboard.
## Watch for Confirm the bet by measuring fewer task retries on form-filling and research jobs. Watch for monthly bills above $50 as the signal to add usage alerts or switch back to a lighter default. Expect a follow-up release that adds cost controls inside the YAML file itself.
Harsh’s take
OpenClaw now pushes users toward a more expensive model by default. The practical result is stronger agent output on real tasks, but solo operators must watch token spend closely or risk surprise bills that wipe out the free-tool advantage.
The trade-off sits between better memory recall across chats and the need to babysit provider limits every month. Many Vibe Builders will see marginal gains on simple automations while paying for capability they rarely tap.
Do this now: add a $30 monthly cap in the Anthropic dashboard and run one full day of scheduled tasks to measure actual cost before committing the default change across all workflows.
by Harsh Desai
About OpenClaw
View the full OpenClaw page →All OpenClaw updatesGo deeper
More AI news
- FeatureHermes Agent verifies work with completion contracts and evidence ledgers
Hermes Agent records verification evidence for coding tasks. The /goal command uses completion contracts to judge success against test runs rather than model assertions.
- FeatureCursor adds cloud agent management to the Agents window
Cursor sets up cloud development environments in under 10 minutes, spins up isolated cloud subagents using /in-cloud, and hands off sessions between local and cloud.
- FeatureCursor introduces /automate skill for automating repetitive tasks
Cursor's new /automate skill creates automations from plain language. Workflows trigger via Slack emojis or GitHub events while cloud agents access virtual computers.