Guide

Give Instinct Your Analytics: A Daily Brief Across Your Sites

Connect Instinct to Agent Analytics with browser approval, then ask for a daily brief about conversions, product use, and what to investigate next.

Give Instinct Your Analytics: A Daily Brief Across Your Sites

“Which site needs my attention today?”

That is a better morning question than checking a traffic chart for every project you run. Connect Instinct to Agent Analytics and ask it to bring you a short daily brief: what changed, which goal it affects, and what to investigate next.

Agent Analytics supplies real product data through its API and CLI. Instinct reads that evidence and handles the recurring task and delivery in your conversation. This workflow uses the official CLI in an Instinct runtime with command execution, persistent storage, and recurring tasks available.

Connect with one browser approval

Send Instinct this prompt:

Connect to Agent Analytics using the official CLI pinned to 0.5.34. Use detached login and send me the browser approval link. Keep the login in persistent private storage for future reporting tasks. Do not ask me to paste an API key.

The command is:

npx --yes @agent-analytics/cli@0.5.34 login --detached

Open the link, sign in with the Google or GitHub identity that owns your projects, and approve access. For the detached flow, send the one-time finish code back to Instinct. It uses the completion command printed by the CLI to exchange the code and save a renewable agent session.

Ask Instinct to verify the connection with an all-sites overview before scheduling anything:

npx --yes @agent-analytics/cli@0.5.34 all-sites --period 7d

You need collected events for a meaningful report. Pageviews establish a baseline; product events such as completed signups, purchases, and useful feature actions let the assistant answer questions about growth. Some deeper analytics require Pro access.

Give the recurring task a clear brief

Copy this prompt and adjust the time and timezone:

Every morning at 09:00 Asia/Jerusalem, check my Agent Analytics projects and send a brief here. Confirm the schedule, timezone, delivery destination, and next run, then run a report now as a check.

Read project goals, activation events, and relevant change annotations. Use existing portfolio context for related projects, while keeping each project's goals separate.

Compare the last completed day with the previous day, and the latest completed seven days with the previous seven. State the exact periods and data timezone. Separate bot traffic where supported. Include counts beside percentages, and flag small samples or missing data.

Tell me what changed, why it matters to my goals, and one useful question or experiment to investigate next. Keep observations separate from possible explanations. If nothing meaningful changed, say so briefly. Report failed reads rather than inventing results. Do not make product or account changes as part of this task.

Connecting the account does not itself create a schedule. Have Instinct confirm that it saved the recurring task and can reuse the login.

Completed windows matter: today’s first nine hours should not be compared with all of yesterday. The all-sites command is a starting overview. For exact comparisons, Instinct can use project queries with explicit date bounds and should check their timezone and boundary behavior.

Give the numbers meaning

A signup flow, a directory, and a documentation site can have different definitions of success. Read project context before interpreting the counts. If goals or important events are missing, ask Instinct to propose them and confirm the initial goals before saving them.

Only group projects into a portfolio when they form a shared growth system. Portfolio context explains their relationship; it does not make every project in your account share one activation definition. Cross-site claims also need the relevant tracking and attribution evidence.

A useful brief might look like this illustrative example, not actual customer data:

Observed: Signup completions fell from 20 to 12 across comparable completed days. Pricing-page visits stayed close to their previous level.

Uncertain: These counts do not tell us whether the cause was the signup form, visitor mix, or ordinary variation.

Next check: Inspect the pricing-to-signup funnel and any recent change annotations before choosing an experiment.

That gives you something to investigate without pretending the data has already established a cause.

Keep access and scheduling separate

The CLI can refresh a valid session. If access expires or is revoked, Instinct should explain the error and ask you to repeat browser approval. Scheduled runs need the same private, persistent CLI configuration used during login.

To stop the daily brief, cancel the task in Instinct. To remove its account access, revoke the session in Agent Analytics Settings / Agent Sessions.

Follow the Instinct connection guide for the full login commands, configuration persistence, and troubleshooting. Then start with one question: what happened across my projects, and what is worth checking next?

Related posts