Parallel web_search with OpenClaw

Page scope: Parallel is a search provider bundled as a plugin—not an LLM. For chat models see Model providers. For page fetching see Firecrawl web_fetch or xAI x_search.

From stable v2026.6.5, OpenClaw ships a bundled Parallel plugin for the web_search tool. Agents can query the web through Parallel's guarded api.parallel.ai/v1/search endpoint instead of wiring a separate skill. Onboarding can discover PARALLEL_API_KEY and surface the provider in the setup picker.

• General web search — News, docs, facts, and research prompts that need fresh results.
• Simple setup — One API key and a bundled plugin; no manual ClawHub skill for basic search.
• Production smoke tests — After upgrade, confirm search before relying on it in cron or channel automations.

Keep specialized tools when you need them: xAI x_search for X/Twitter context (migration guide), Firecrawl for structured page fetch (web_fetch migration), or skills from ClawHub for niche sources.

• OpenClaw v2026.6.5+ (or current beta if you are explicitly testing pre-release builds).
• A Parallel API key from parallel.ai (or your Parallel dashboard).
• Node.js 22.19+ and a running Gateway after config changes.

1. Upgrade if needed: openclaw update then openclaw doctor --fix.
2. Set PARALLEL_API_KEY in your environment or secrets store (Secrets).
3. Run openclaw onboard or use the Setup Wizard—the Parallel provider should appear in the web-search picker on supported versions.
4. Restart the Gateway: openclaw gateway restart.
5. Ask the agent a search-heavy prompt (e.g. “search for recent OpenClaw release notes”) and confirm tool calls succeed.

export PARALLEL_API_KEY="your-key-here"
openclaw gateway restart

Production operators: add Parallel to the smoke list on Deployment Upgrade Runbook → v2026.6.5+.

openclaw doctor
openclaw gateway restart