Tool Question Answer
Category: Site Knowledge · Since: v18.60.0 · Tier: All
Ask any question about CBrowser and get an answer grounded in its documentation, with the sources it used.
Since 20.0 it works everywhere the tool runs, including the local npx cbrowser mcp-server. The CLI's cbrowser ask uses the same client.
What It Does
This tool queries the CBrowser knowledge base when you need to understand:
- What a tool does and how to use it
- What a score or metric means
- Which persona to use for a specific scenario
- How tools work together in a workflow
- What a trait value means at a specific level
The answer is written from the documentation the search finds, and the response lists those sources so you can read them.
Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
question |
string | Yes | Your question about CBrowser |
context |
string | No | URL being tested, tool in use, or scores you see |
Where the question goes
| Where the tool runs | Who answers |
|---|---|
Your machine (npx cbrowser mcp-server) |
cbrowser.ai, anonymously. Your question is sent there. CBROWSER_HELP_URL points it somewhere else |
| The hosted servers (demo, pro, enterprise) | The hosted documentation service, with your account. cbrowser.ai only if that service cannot be reached |
A server with CMS_URL set |
Only that service |
Your cbrowser key is never sent to the public endpoint at cbrowser.ai. A request gives up after 30 seconds. The public endpoint is rate-limited. An HTTP error such as a 429 or a 500 comes back as a tool error and is not retried elsewhere, so a question is never paid for twice. Before 20.0 the tool only worked on the hosted servers and failed on a local server.
Examples
question_answer(
question:
"What does empathy_audit do?"
)
question_answer(
question:
"How do I test on mobile?"
)
question_answer(
question:
"What does a CTC of 0.85 mean?",
context:
"Readability is the bottleneck"
)
question_answer(
question:
"Which persona fits seniors?"
)
Response
Abridged:
{
"question":
"What does empathy_audit ...",
"answer":
"empathy_audit simulates ...",
"sources": [
"[/docs/tool-empathy-...]: ...",
"[/accessibility ...]: ..."
],
"note":
"Answer grounded in the ..."
}
A request that fails (rate-limited, unreachable, timed out) is a tool error whose error says what happened.
How It Works
- Your question is sent to the CBrowser knowledge base
- Full-text search finds the most relevant documentation pages
- Tool names in your question trigger extra lookups for those tools
- Claude reads the matching docs and answers based only on what they say
- Source references are returned so you can read the full docs
Knowledge Base
The knowledge base indexes the pages of cbrowser.ai, these docs included, and the cognitive trait reference. It is rebuilt when the site is deployed.
When to Use
Use question_answer when you:
- Are unsure which tool to run next
- Need to understand what a score means
- Want to know which persona fits a scenario
- Need to understand how two tools differ
- Want to confirm a tool accepts a specific parameter
Tips
- Be specific. "What does CTC measure?" gets a better answer than "explain scores"
- Include context. Mentioning the URL or tool you're using helps the answer focus
- Ask about workflows. "How do I do a full mobile WCAG audit?" returns multi-tool guidance
Related Tools
page_understand— Analyze a page's structure and affordancessite_model_query— Query a learned site modellist_cognitive_personas— List available personas