OpenClaw installation and usage

Camofox Web Search integrates with OpenClaw through a native npm plugin. OpenClaw continues to expose its canonical web_search and web_fetch tools; only the backing providers change to your self-hosted gateway.

Prerequisites

• A running Camofox Web Search deployment and its public WEB_SEARCH_API_KEY.
• OpenClaw 2026.7.1 or newer, with a working model provider.
• Node.js 22 or newer for the camofox-web-search installer.

Verify the gateway before changing OpenClaw:

curl --fail https://search.example.com/healthz
npm install -g camofox-web-search

Managed installation

Export the key for the installer and diagnostics, then install the native provider:

export WEB_SEARCH_API_KEY="<copy securely from the server .env>"

camofox-web-search install openclaw \
  --endpoint https://search.example.com \
  --scope user

The installer:

• installs camofox-web-search-openclaw;
• configures camofox as OpenClaw's search and fetch provider;
• stores an environment SecretRef, not the token value;
• preserves unrelated OpenClaw settings and records the previous providers for uninstall.

If another provider already owns either capability, inspect the conflict first and rerun with --force only when replacing it is intentional.

Make the key available to the Gateway

An OpenClaw Gateway running under systemd, launchd, or another supervisor does not inherit variables exported only in an interactive shell. Put the key in OpenClaw's runtime environment file:

mkdir -p ~/.openclaw
chmod 700 ~/.openclaw
touch ~/.openclaw/.env
chmod 600 ~/.openclaw/.env

Add this line to ~/.openclaw/.env without committing the file:

WEB_SEARCH_API_KEY=<your-public-gateway-key>

Restart and verify the runtime registration:

openclaw config validate
openclaw gateway restart
openclaw gateway status
openclaw plugins inspect camofox --runtime --json

camofox-web-search doctor openclaw \
  --endpoint https://search.example.com \
  --scope user --live

doctor checks the local config, gateway health, MCP contract, native OpenClaw search/fetch provider registration, and—when --live is present—a real search.

Use it

Start the OpenClaw TUI and ask for a search:

openclaw tui

Example prompt:

Use web_search to find the latest Camofox Browser documentation, fetch the primary source, and cite it.

For a one-shot verification after configuring an OpenClaw model:

openclaw agent --agent main \
  --message "Use web_search to find the OpenAI official site and return its first URL."

Proxy and local-model deployments

When OpenClaw uses HTTP_PROXY or HTTPS_PROXY, exclude the Web Search gateway and any local model endpoint from proxying. For a Lima VM that reaches an oMLX server on the host, for example:

NO_PROXY=127.0.0.1,localhost,host.lima.internal,*.local,192.168.*
no_proxy=127.0.0.1,localhost,host.lima.internal,*.local,192.168.*

Restart the Gateway after changing .env. A Gateway that is reachable while model calls fail with fetch failed usually indicates a missing local-model host in NO_PROXY.

Manual installation and removal

The checked-in OpenClaw example contains the equivalent manual plugin and JSON5 configuration.

Remove only the managed integration and restore the providers that were present before installation:

camofox-web-search uninstall openclaw \
  --endpoint https://search.example.com \
  --scope user
openclaw gateway restart

Removing the integration does not delete WEB_SEARCH_API_KEY from your secret environment file.