What does web search add to an OpenClaw agent bill?

Keep two ledgers, then reconcile them

If your OpenClaw agent uses web search, a token-only estimate can miss a search-provider charge. It can also overstate that charge if you price every tool invocation as a fresh paid request. Start with the model's billable token usage. Put externally billed search requests in a second ledger, then compare both against the statements for the accounts that actually served them.

Find out which search route actually ran

OpenClaw's web-search documentation describes configurable managed providers, key-free options, and a query cache whose default duration is 15 minutes. A direct OpenAI Responses model can use its native web search instead when no managed provider is pinned. That makes the configured provider, model route, and account more useful for billing than the tool name alone. Check the effective route in OpenClaw's Search settings and the relevant provider usage record; do not paste credentials into a cost worksheet.

A tool invocation, a search served from cache, and a provider-billed request are different events. A retry may add another request. A native model-provider search may have different billing rules from a separate Search API. If you cannot match requests to a provider statement, keep them in an unknown bucket rather than assign a price.

A request worksheet you can reuse

For one billing period	Where to check	What to count
Model input and output	Model-provider usage and invoice	Billable tokens, including any separately priced cache category
Managed search	Effective search route, search-provider usage and invoice	Provider-billed requests and applicable plan credits
Native or key-free search	Provider terms and the account statement	Any separate tool fee, or explicitly unknown if not reported

For a worked illustration, suppose an agent invokes search 120 times in a month, but the search-provider account reports 80 billable requests. On October 14, 2026, Brave's Search API page lists $5 per 1,000 requests for its Search plan and $5 in monthly credits. At that listed rate, 80 requests would be $0.40 before credits, taxes, and any other plan terms. The 120 and 80 are hypothetical, not Clawback measurements; the difference is not proof that 40 calls hit a cache. Use your actual provider count and current terms instead. Do not apply Brave's rate to native search or another provider.

What Clawback's calculator covers

The free Clawback calculator estimates 30 days of model-token spend from selected models, messages, heartbeats, sub-agents, and channels. Its implementation has no web-search request field and does not import an invoice. Run a model-token scenario there if that helps you budget, then add a separately verified search line outside the tool. If the model provider already includes a search charge on the same statement, reconcile that line rather than add a second assumed fee. The final number is an estimate until your model and search-provider statements agree on dates, accounts, and billing units.