Reader Questions - Docsalot

Ask your AI assistant what readers need help with, explore recurring topics, and turn the findings into documentation improvements.

Reader Questions works through Authoring MCP. It requires an Enterprise plan, access to the documentation, and the analytics:read permission.

Enable analytics

Connect with OAuth

Connect Authoring MCP

Follow the setup steps for your client.

Approve analytics access

Select Inspect Reader Questions analytics when prompted. If you connected without this permission, reconnect and authorize again.

Connect with an API token

Issue a token

In your Docsalot team’s Settings → API Tokens, choose Issue a new token.

Enable Reader Questions

Select Allow Reader Questions analytics, then create and copy the token.

Reconnect your client

Replace the token in your MCP client’s configuration and reconnect. Existing tokens do not gain analytics access automatically.

Ask about your readers

Start with a project and a reporting period:

For "My Project", summarize what readers asked last month. Include the date range, top themes, and number of contributing sessions.

Then investigate a topic:

Show up to five representative samples about MCP setup. Explain what readers struggled with and separate the evidence from your interpretation.

Use the findings to improve a page:

Based on these samples, suggest three documentation improvements. Draft the most useful change and save it without publishing.

For a comparison, ask for the last 30 days versus the previous 30 days. For a calendar month, specify its name or exact dates.

Understand the results

The overview includes signal counts, contributing sessions, and top themes. Ask for trends or recommendations to explore further, and request samples to see the supporting evidence.

Samples are stored summaries, not verbatim user questions or complete assistant responses. They may summarize an assistant’s answer rather than a reader’s question. A signal count is not a count of unique questions or readers.

Use summaries to identify topics to investigate. Check claims against your documentation before treating them as product facts, and avoid broad conclusions from a small number of sessions. Samples omit customer, session, and internal memory identifiers.

Available tools

get_reader_question_insights retrieves one view for a documentation project.

Start with view: "overview".
Use themes, trend, or recommendations to go deeper.
Choose a range: 7d, 30d, 90d, 12m, or all.
For specific dates, supply both from and to as YYYY-MM-DD in UTC.
Use compareWithPrevious to compare with the preceding equal-length period.
Filter with topic and request up to five themes or recommendations with limit.

Use get_reader_question_insights for <DOCUMENTATION_ID> with range "30d", view "overview", and compareWithPrevious true.

list_reader_question_samples retrieves up to 10 summaries at a time.

Use the same project and date range as the overview. Filter by topic or kind (user_summary, coverage_gap, or feature_summary). If nextCursor is returned, pass it as cursor to get the next page with the same filters.

Use list_reader_question_samples for <DOCUMENTATION_ID> with range "30d", topic "MCP setup", and limit 5.

Your assistant can get the documentation ID with list_documentations. You do not need to look it up before asking a question.

Troubleshooting

For connection help, see Authoring MCP. For other issues, contact faizank@docsalot.dev.

​Enable analytics

​Connect with OAuth

Connect Authoring MCP

Approve analytics access

​Connect with an API token

Issue a token

Enable Reader Questions

Reconnect your client

​Ask about your readers

​Understand the results

​Available tools

​Troubleshooting