How to use Hamster: Small teams

A guide for small teams (1–10 people) shipping daily. Often founder-led, engineers already on Cursor or Claude Code. We'll start by connecting your tools so the AI has grounding, then walk through shipping your first brief, the IDE flow for engineers, Cloud Agents, and the Blueprint that comes out of it.

Small teams overview

~2 minutes

Video coming soon

Connect your tools

Hamster's value compounds with how much of your team's work the AI can see. The Context Graph is the connected, summarised view of your code, docs, tickets, designs, and conversations — and it's what grounds every Brief, Plan, and chat reply in your real systems instead of generic AI guesses.

Three connections do the most work on a small team:

  • GitHub — required for delivery, and the largest single source of code-level context. Hamster reads your repos, PRs, and issues.
  • Google Drive — turns your existing specs, wikis, and shared docs into context the AI reads.
  • Linear — bidirectional sync of briefs, tasks, and initiatives, so existing rituals (stand-ups, sprint review) stay intact.

Add the rest as you grow into them: Slack for notifications and the brief side panel, Figma for design context, Notion for wiki pages, Meeting Agent for customer calls. See Connections overview for the full list.

Ship your first Brief

Open a chat with Hamster and describe what you want to build. The AI assistant asks the right clarifying questions, pulls in any context you drop in, and produces a structured Brief.

Discovery doesn't have to start in a Brief either — spike code in your IDE to test a hunch, ideate in Figma, whiteboard with your team, then come back when the shape is clear and ask Hamster to turn the conversation into a Brief.

A Brief captures the goal, the user need, the scope, the success criteria, and the context an engineer or AI agent needs to ship the work. It's not a PRD. It's not a ticket. It's the artefact your team and the AI both read from. See Creating Briefs for the full surface.

Examples of Briefs:

  • "Add tag-based filtering to the briefs list view"
  • "Investigate the spike in churn for the trial cohort and recommend three changes"
  • "Build a public share page for individual customer success stories"

Generate a Plan

Once the Brief looks right, generate a Plan. The Plan is a structured task breakdown derived from the Brief — mostly internal scaffolding for the AI agent. Glance at it before delivery to make sure the agent understood the scope; don't manage it the way a PM would in a project tool.

Top tip: If the Plan misses something, ask in the chat. Hamster regenerates the Plan from your follow-up — no need to edit it by hand.

Deliver

Click Deliver. The Cloud Agent picks up the Plan, executes it, and pushes a PR to GitHub. The delivery thread streams the agent's narrative live — every tool call, every flow step, the PR push — so you see what's happening and can interrupt if the agent drifts.

Code review still happens on GitHub the way you already do it. Hamster doesn't replace your review process; it produces the PR for you to review.

Wire up the IDE flow for engineers

Your engineers already use Claude Code, Cursor, Codex, or another AI coding tool. The fastest way to make Hamster useful for them is to expose your Context Graph, Briefs, Blueprints, and Methods to whatever they already use.

Two pieces:

  • The Hamster CLI — installs in their terminal. Sync skills, run slash commands, scope context to a brief.
  • The MCP server — once connected to Claude Code or Cursor, the engineer can say "ship the [brief title] brief" and the agent has the brief plus all its grounding context immediately. No copy-paste.

This is the highest-leverage thing you can do for engineer adoption. Engineers stay in control, ship from where they live, and get the team's context for free.

Configure one Cloud Agent per repo (optional)

A Cloud Agent is the configured environment Hamster runs deliveries in when nobody's hands are on a keyboard. It knows your repo URL, your env vars, your build and test commands.

Cloud Agents shine when a PM clicks Deliver on a Brief without an engineer in the loop, when you want parallel deliveries across the team, or for background work the team doesn't want sitting on someone's laptop. Engineers usually ship from their IDE, not from Cloud Agents — see the previous section.

If you do want Cloud Agents: configure one per active product repo. Paste your .env, hit Save, and Hamster auto-builds and snapshots the environment. Status flips to Active when it's ready, usually in under a minute.

Top tip: If your .env changes, hit Reset on the Cloud Agent. Hamster rebuilds in place — you don't reconfigure from scratch.

Let Hamster write your Blueprint

Once your Context Graph is connected, Hamster generates an initial Blueprint — an English representation of what your product is today, derived from your code, your existing docs, and your tickets. This is the source of truth every Brief gets refined against.

Spend an afternoon reviewing and editing what Hamster generated. Things to look for:

  • Product overview — does this describe what the product is and who it's for, the way you'd describe it to a new hire?
  • Architecture — are the major systems and how they fit together captured accurately?
  • Constraints — anything explicit you'd want a new engineer (or an AI agent) to know before touching the code: testing patterns, deploy procedures, data-handling rules, security gates.

You're editing what Hamster proposed and adding what only you and the team know. The Blueprint is bidirectional — drop in additional docs (a Notion page, a tech spec) and it updates; edit it directly and downstream artefacts re-derive.

Top tip: Have everyone on the team read the Blueprint once it's settled. It's a good check on shared understanding — surfaces where the team has been operating on different mental models.

When a Brief lands and the PR merges, the Context Graph propagates the change back into the Blueprint — surfacing a "this part may need updating" prompt for review, or applying obvious updates automatically. Change ships, what's true updates, the Blueprint stays current.

Set Direction with Goals and Initiatives (when you're ready)

Naming the outcome before the work isn't required to ship. Once you've delivered a few Briefs, though, the Goals and Initiatives layers stop tickets piling up.

Pick a framework your team already uses. For most small teams, OKR (Objective → 2–5 Key Results) or North Star (one primary metric → 3–5 inputs) are the natural starting points. Don't pick more than one on a small team — pick one and use it.

Attach a metric to each measurable Goal — unit, direction, target — so the team and the AI trade in the same numbers. Log results per period as you go.

Initiatives commit to a Goal. Examples for a small team:

  • "Ship the v1 onboarding flow" → Activation Goal
  • "Validate the pricing-page experiment" → Revenue Goal
  • "First-paying-customer push for cohort A" → New Logos Goal

When you write a Brief, attach it to the right Initiative. The Initiative gives you a single status surface for everything in flight that ladders up to that Goal — useful even at three people, because it stops you optimising for "tickets shipped" and points the team at the outcome.

Methods at this stage

For small teams, the default Hamster Method covers most patterns out of the box. You don't need to fork or write custom Methods until you spot a recurring kind of work where the AI is missing your team's specific way of doing it. When that happens, fork the Hamster Method into your workspace and add what's needed.

Understanding progress

The activity timeline

Every brief has an activity timeline that records every change — human or AI — to the brief and its plan. Alignment votes, status changes, scope edits, agent runs. It replaces "what's the latest?" Slack pings.

Slack as your status surface

For non-builders on the team (a designer, a co-founder, a sales partner), the Slack brief side panel is usually the right surface. It renders the live brief inline in Slack so they don't need to log into Hamster to know what's going on.

Going deeper

When you've shipped 10–20 briefs and want to level up:

  • Routines — automate repeating work like post-delivery summaries and plan reviews.
  • Methods Library — fork the Hamster Method and customise for your team.
  • MCP server — expose Hamster's context to Cursor, Claude Code, and other tools.

When you outgrow this guide, see How to use Hamster: Startups & mid-size companies.