---
title: your agency's direction, inside your coding agent
description: Atlas by Friendlym keeps a business dossier for each of your software projects, turns it into tasks from your agency, and lets your own coding agent read those tasks, ask for direction and report progress back.
canonical: https://tilqui-atlas.pages.dev/?lang=en
lang: en
last_updated: 2026-10-08
alternate_html: https://tilqui-atlas.pages.dev/?lang=en
---

# Atlas by Friendlym

Atlas by Friendlym turns a pile of software projects into a short list of what to do next. Your agency keeps a business dossier for each project you share: the market, the customers, the competitors, pricing, risks and how close it is to launch. From that it assigns you tasks. Your own coding agent (Claude Code, Codex, Cursor or any MCP client) reads those tasks, asks Atlas for direction and posts your progress back, so the agency always knows where things stand.

## What it does

- **A dossier per project.** Research on the business side of each project, with every claim labelled as verified, sourced, estimated or assumed, and a score and verdict you can argue with.
- **Tasks from your agency.** "Your billing project has potential: ship the payment page, then talk to five customers." Tasks point at the evidence behind them.
- **Your agent, connected.** Ask "what should I do?" in your own coding agent. It signs in to Atlas with standard OAuth, reads your tasks, asks Atlas for direction and tells you what to do next, with references.
- **Progress the agency sees.** When you or your agent finish a step or get stuck, the agent posts a note on the task. The agency sees it on their side and answers.

## How onboarding works

Atlas never touches your computer. Your own agent, which you already trust, does the reading, and you see exactly what Atlas receives.

1. **Copy a prompt.** After your invite, the welcome page shows one prompt as plain text. Paste it into your coding agent. It looks through your code folders and notes, for each project, the name, stack, git activity, a README summary, live URLs and one line on what it is for. It never opens `.env` files, keys or source code beyond READMEs and package manifests, and it sends nothing.
2. **Read the file.** The agent writes everything into one file, `atlas-census.md`. It is a normal document: open it, edit it, delete the projects you want to keep private, mark the important ones.
3. **Upload it.** Drag the file onto the Atlas page, or let your agent send it with a one-time upload key that can upload one file and read nothing. Atlas checks it for anything that looks like a secret before it accepts it.

Then Atlas triages your projects, you and the agency agree on the ones that matter, and research runs on those. Your site, chat and tasks are ready when it is done.

## How agents connect

Agents use the same standard sign in as any modern API: OAuth 2.1 with PKCE, dynamic client registration, and the device flow for agents without a browser. You approve each agent on a consent screen that shows its name, what it may do and in which workspace, and you can remove it at any time.

```sh
claude mcp add --transport http atlas https://tilqui-atlas.pages.dev/mcp
codex mcp add atlas --url https://tilqui-atlas.pages.dev/mcp
```

Step by step for Claude Code, Codex and Cursor: [MCP setup](https://tilqui-atlas.pages.dev/docs/mcp/?lang=en). The sign in in detail: [authentication](https://tilqui-atlas.pages.dev/docs/auth/?lang=en). Everything at once: [docs](https://tilqui-atlas.pages.dev/docs/?lang=en).

## What stays private

Everything about your projects (dossiers, scores, tasks, notes) is private to your workspace and needs a sign in. These public pages only describe the product. Atlas stores summaries your agent wrote, never your source code.