# Scout, the AI that reads all of it and asks before it acts

Scout compares your numbers across every source, explains what may be in the way, prepares the change and checks whether it helped. It remembers what worked and works with your own agents through MCP.

[Open Scout AI in the demo workspace](/tractionscout/demo/workspace?view=scout) (sample data, no account)

## What you can do

- **Evidence first:** Scout separates what the data shows from what still needs checking, and says when a finished task is not yet proof of more customers.
- **Prepares the work:** Copy, content, a code change, a post or a campaign, ready for you to review in Work.
- **Separate permissions:** Reading data, preparing a draft, publishing and changing a budget are separate permissions. Approvals stay with you.
- **Memory you control:** Scout remembers what worked per project, from experiments, assets, posts and alerts. Review, approve and retire memories in Activity & Memory.
- **Automations:** Work Scout repeats hourly, every 6 or 12 hours, daily or weekly, with a record of what came of each run.
- **MCP and Agent CLI:** Connect Claude Code, Codex or any MCP client with the open-source Agent CLI. Read-only by default, and external execution is always blocked.

## What Scout does

Scout is the AI inside TractionScout. It reads every source you connect, from analytics and search to payments, social and email, and works through a growth question the way a careful analyst would:

1. compare the numbers along your funnel,
2. explain what may be getting in the way, keeping what the data shows apart from what still needs checking,
3. prepare a change: copy, content, a code change, a post or a campaign,
4. check afterwards whether it helped.

A finished task is not treated as proof of more customers or revenue. Scout says when the evidence is not in yet.

## You stay in control

Reading data, preparing a draft, publishing and changing a budget are separate permissions. Everything Scout prepares appears in **Work**, where you review, approve and follow each growth action. Reading your analytics never gives Scout permission to post or spend.

## Memory that you curate

Scout keeps a memory per project, built from evidence such as experiments, assets, social posts and alerts. In **Activity & Memory** you can review, approve and retire what it remembers, see what it made, and check the health of scheduled work.

## Automations

Automations are work Scout repeats on a schedule, hourly, every 6 or 12 hours, daily or weekly, with a record of what came of each run. You choose what happens when a previous run has not finished. Scout's notes can reach you as browser notifications.

## Bring your own agents with MCP

The open-source Agent CLI (`tscout`, AGPL-3.0) connects Claude Code, Codex or any MCP client to your projects. The download is verified with a checksum and has no dependencies. The default mode is read-only; a review-staging mode lets agents stage work for your approval. External execution is blocked in every mode, and credentials, approvals and budgets stay on the server.

Set it up with [Agent CLI and MCP](/docs/agent-cli.html), or see an example in [Connect a coding agent or script](/workflows/agent-cli-automation.html).

## Questions

### Will Scout post or spend money on its own?

No. Reading your analytics does not give it permission to publish content or change a budget. Those actions need the right connection and follow the approval rules you set.

### Can I use my own AI agent?

Yes. Install the Agent CLI and connect any MCP client. It works with the same projects and permissions, while credentials, approvals and budgets stay on the server.

### Which AI model does Scout use?

The model and AI budget are set per project in AI & Billing settings. Provider usage may cost extra depending on your plan.


Maintained by TractionScout · Updated 2026-11-05


## Setup guides

- [What is TractionScout?](https://tractionscout.com/docs/overview.html)
- [Agent CLI and MCP](https://tractionscout.com/docs/agent-cli.html)
- [Managed service, pricing, and privacy](https://tractionscout.com/docs/managed-service.html)

## Start with this question

What is holding back sign-ups this month, and what should I try first?

[undefined](https://tractionscout.com/signup/tractionscout?prompt=What+is+holding+back+sign-ups+this+month%2C+and+what+should+I+try+first%3F)

Your question is carried into setup. Review it, connect the data you need, and send it to Scout when you are ready.

## Related examples and guides

- [KPIs that adapt to what each product is trying to achieve](https://tractionscout.com/features/smart-kpis.md): Each project’s board leads with its objective and picks the figure that fits each growth step from the sources you connected, from active users to trial-to-paid, without a spreadsheet of metric definitions.
- [See how you are found in Google, Bing and AI answers](https://tractionscout.com/features/seo-ai-search-visibility.md): Search Console and Bing data, rank tracking, keyword ideas, site health and the answers ChatGPT, Claude, Gemini and Perplexity give about your market, all tied to the sign-ups search brings.
- [Plan, cross-post and measure social posts in one calendar](https://tractionscout.com/features/social-media-scheduling.md): Write once, adapt per network, schedule into one calendar and see which posts brought visits and sign-ups, with posting times suggested from your own results.
- [An AI support chat that answers from your own pages and hands the rest to you](https://tractionscout.com/features/ai-support-chat.md): Scout reads your site and help pages, answers visitors with cited replies, and turns anything it should not answer into a case in your Work queue, with a draft reply ready.
