How to use Hamster: Startups and mid-size companies

A guide for admins setting up Hamster at a 10–50 person product org. Post-PMF, multiple product teams, engineers all using AI coding tools. Covers connecting a richer Context Graph, setting Direction with Goals, shaping the Initiatives that move them, and steering the Briefs Hamster drafts beneath each Initiative.

Under 10 people? How to use Hamster: Small teams is the right starting point. Want to try Hamster on a single Brief first without the setup? See Just ship.

Running Hamster across multiple teams

~3 minutes

Video coming soon

Setting up your workspace

Organise your teams

At this stage you typically have one Hamster workspace with multiple teams underneath — for example, "Growth", "Core Platform", "Mobile", "Marketing". Each team has its own Briefs, Plans, and deliveries; integrations and Cloud Agents live at the workspace level.

Walk through onboarding once with the founding admin, then invite the rest of the org. Default new invites to Reviewer and promote to Creator as people start writing briefs. The Reviewer role is read-only on Briefs and Plans but can comment, vote alignment, and use Slack — a good default for non-builders and stakeholders. See Roles & permissions for the full role model.

Top tip: Send the Joining a team on Hamster guide to everyone you invite. It gets ICs (engineers, designers, stakeholders) oriented in 10 minutes.

Connect your full Context Graph

A richer Context Graph is the single biggest difference between this stage and the small-team setup. The more your team has accumulated in real systems — Linear tickets, Figma frames, customer-call recordings — the more grounded the Blueprint Hamster generates and the better every Brief gets refined.

Connections worth investing in here:

  • GitHub — at this scale, configure at the GitHub-org level. Hamster reads code and PR history across all your repos.
  • Linear (or Jira) — bidirectional sync of Briefs, Tasks, and initiatives. Briefs in Hamster, task-level state visible in Linear, so existing rituals (stand-ups, sprint review) stay intact.
  • Figma — designers attach .fig files and Figma URLs to Briefs; the AI agent uses them as grounding context for the Plan and delivery.
  • Notion and Google Drive — turn existing wiki pages and shared docs into context the AI agent reads.
  • Slack — the bot, the Brief side panel, slash commands. Where most non-builders interact with Hamster.

See Connections overview for the full list.

Wire up the IDE flow for engineers (the most-used path)

At this scale, most of your engineers already use Claude Code, Cursor, Codex, or similar. The biggest unlock for adoption is exposing your Context Graph, Briefs, Blueprints, and Methods inside the tool they already use.

Two pieces, installed by each engineer:

  • The Hamster CLI — handles auth, skills sync, slash commands, scoped context. Lives in the terminal next to git and gh.
  • The MCP Server — once connected to Claude Code or Cursor, an engineer can say "ship the [Brief title] Brief" or "show me what's in [Initiative]" and their AI assistant has every piece of context Hamster's seen. No copy-paste, no tab-switching.

This is the path that makes engineers happy. They stay in control, ship from where they live, and get the team's accumulated context for free. Make sure the MCP setup is documented in your engineering onboarding.

Cloud Agents per repo (for delivery without an engineer at the keyboard)

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

Cloud Agents complement the IDE flow rather than replacing it. They're how:

  • A PM kicks off a Delivery on a Brief without an engineer in the loop.
  • Routines run scheduled or event-driven deliveries.
  • Multiple deliveries happen in parallel without queuing on engineer time.

If your team has multiple repos shared across product teams, configure one Cloud Agent per repo. If you have multiple environments (e.g. staging-api and production-api), configure a Cloud Agent per environment.

Top tip: Lock production-deploying Cloud Agents to senior engineers using the permissions model. Reviewers can read configurations but can't trigger deliveries.

Building blueprints per product surface

At small-team scale, one Blueprint covers most of the product. At 10–50 people, you'll typically have multiple product surfaces — a web app, a mobile app, a public API, an internal admin tool — and each deserves its own Blueprint.

Hamster generates first drafts from your Context Graph. Spend a sprint reviewing and editing them. The pattern that works: each product team owns the Blueprint for the surface they ship into, edits it as the surface evolves, and treats the Blueprint as the canonical answer to "what is this surface today?"

Blueprints are bidirectional. If your team migrates from one architecture to another, the Blueprint updates as the new code lands. If you author a new architecture doc in Notion, the Blueprint absorbs it on the next sync.

Anchor work in Goals

At 10–50 people, "what are we optimising for this quarter?" stops being obvious. Goals is where the answer lives, in a framework your team already speaks.

Use the language and framework your team already uses. Hamster does not require you to adopt a new one before work begins.

Attach a Metric to each measurable Goal — unit, Direction, target, optional baseline. Log Results per period; status (On Track, At Risk, Off Track) makes "off track" visible without anyone curating a deck.

Shape Initiatives from those Goals

For each Goal, shape the Initiatives that could move it — a quarterly bet, a launch, or a discrete strategic project. Each Initiative links to one or more Goals, with optional weights, and becomes the middle layer between the outcome and the work.

Initiatives matter at this scale because:

  • PMs talk to your CEO about outcomes, not tickets.
  • Cross-team dependencies are easier to express ("this Initiative depends on that Initiative") than at the Brief level.
  • Status roll-ups happen at the right level of detail for execs.

Examples of Initiatives:

  • "Mobile app launch — Q3" → Mobile adoption Goal
  • "Self-serve checkout" → Activation Goal
  • "Sales-led onboarding for top 100 accounts" → Enterprise revenue Goal

The pattern is Goal → Initiative → Brief → Plan. Once the team agrees on an Initiative, Hamster drafts each related Brief. PMs, leads, designers, and engineers steer the Brief, then Hamster generates its Plan; execs can stay at the Goal and Initiative level rather than reading every Brief. If you have Linear or Jira connected, Initiatives sync to your existing exec dashboards.

Shipping change

The Discovery → refinement → PlanDelivery loop is the same as it is for small teams. Spike code, ideate in Figma, run a Research Agent pass, or whiteboard with your team; bring what you learn back to the Initiative and its related Brief. Each Brief aligns the team on one deliverable, then feeds its Plan.

What changes at this scale:

  • Run Delivery from Briefs, not terminals. Engineers might still kick off the occasional CLI Delivery, but the primary surface is the Brief — every Delivery is attributed, observable in the delivery thread, and reviewable by anyone in the org.
  • Alignment becomes asynchronous. At small scale, alignment happens in the room. At 30 people it happens in the activity timeline and Slack alignment unfurls. Make sure every Brief has a clear set of alignment voters.
  • Multiple teams ship in parallel. Cloud Agents per repo means deliveries don't queue behind each other.

Customize Hamster with Methods

Hamster is the harness and automatically applies relevant Methods while shaping Briefs, generating Plans, and delivering work. Methods are optional components for customizing that harness with your team's conventions; they are not a setup gate.

When shared conventions emerge across teams, add them at the workspace level rather than maintaining a separate set per team.

Common additions teams make at this scale:

  • Testing patterns — how the team expects unit, integration, and E2E tests for different kinds of change.
  • Deploy procedures — pre-merge checks, snapshot environments, manual approval gates.
  • Design-system compliance — when a Brief touches design-critical surfaces, what checks are required before the PR is reviewed.
  • Domain-specific patterns — anything specific to your product that the AI shouldn't have to re-figure-out per brief.

Top tip: Add "How we test" and "How we deploy" Methods when those conventions need to differ from Hamster's defaults.

Understanding progress

Initiative roll-ups, not brief lists

Once you have multiple teams writing Briefs, the Brief list view is too noisy for status. The Initiatives view and the per-initiative timeline are the right surfaces for execs and stakeholders.

Activity timelines on briefs

Every Brief has an activity timeline — every change to the Brief, every alignment vote, every Plan revision (including AI-driven revisions), every Delivery run. It's the audit log for one piece of work.

Slack for stakeholders

For non-builder stakeholders (sales, CS, ops, designers who aren't in Hamster every day), the Slack brief side panel renders the live Brief inline. They don't need to log into Hamster to stay in the loop.

Going deeper

  • Routines — automate "every time X happens, do Y" patterns.
  • MCP Server — expose Hamster's Context Graph to Cursor, Claude Code, and other tools your engineers use.

When you cross 50 people or start running multiple product orgs, see How to use Hamster: Enterprise and scaling.