Problem

browser tool has a configurable ssrfPolicy in openclaw.json:

"browser": {
  "ssrfPolicy": {
    "allowPrivateNetwork": true,
    "allowRfc2544BenchmarkRange": true,
    "allowedHostnames": ["example.local"]
  }
}

However, web_fetch has no equivalent config — it hardcodes strict SSRF mode via withStrictGuardedFetchMode(), which passes no policy to fetchWithSsrFGuard().

Use Case

Home network with Surge (macOS proxy) acting as DHCP + DNS for all LAN devices. Surge uses fake-IP mode, resolving all domains to the RFC 2544 benchmark range (198.18.0.0/15). This causes web_fetch to block every URL that goes through Surge DNS, even for legitimate public websites.

The browser tool works fine because it has ssrfPolicy.dangerouslyAllowPrivateNetwork / allowRfc2544BenchmarkRange config options. But web_fetch has no such config, forcing users to patch source code.

Current Workaround

Patching withStrictGuardedFetchMode() in all bundled JS files to inject allowRfc2544BenchmarkRange: true into the policy. This needs to be re-applied after every openclaw update.

Proposed Solution

Add ssrfPolicy config under tools.web.fetch, mirroring the browser config schema:

"tools": {
  "web": {
    "fetch": {
      "ssrfPolicy": {
        "allowRfc2544BenchmarkRange": true
      }
    }
  }
}

The implementation would pass this policy through fetchWithWebToolsNetworkGuard() → fetchWithSsrFGuard() → resolvePinnedHostnameWithPolicy(), which already supports the full policy object.

Environment

• OpenClaw 2026.3.13
• Home LAN with Surge proxy (fake-IP mode, 198.18.0.0/15)
• Linux host (gateway mode: local)