Getting started

From sign-in to your first reviewed pull request. Most teams get here in one sitting.

  1. 1

    Sign in and pick your workspace

    Sign in with GitHub or Google. A workspace belongs to your team or company — it holds teammates, projects, integrations, and billing. With GitHub sign-in you’ll pick which personal account or organization to onboard; with Google you get a personal workspace and can connect GitHub later.

  2. 2

    Install the GitHub App

    Gentlybot talks to your repositories through a GitHub App you install on the account you chose. You control which repositories it can see, and you can revoke the installation at any time. Gentlybot doesn’t keep a copy of your code — it clones on demand into a sandbox for each run.

  3. 3

    Create your first project

    A project groups the repositories that make up one product — the primary repo plus any context repos (a docs repo, a sister service). Name it and select repositories; that’s the context threads will work against.

  4. 4

    Teach Gentlybot to run your app

    Sandboxes boot your real stack from an apps.yml config: the toolchain, services like Postgres or Redis, and each app’s setup, startup, ports, and health checks. Project settings hold the shared baseline; thread agents use workspace-only overrides while testing branch-specific code.

    apps.yml reference
  5. 5

    Start your first thread

    Ask something real: how a flow works, whether a request is feasible, or go straight to a change. Threads run in a mode — Auto routes for you; Ask answers with citations and touches nothing; Plan drafts an approach you approve; Implement builds, verifies in a live sandbox, and ends in a pull request your team reviews.

  6. 6

    Invite the team and connect tools

    Invite unlimited teammates — pricing is per workspace, not per seat. Connect Slack, Jira, Notion, or your own MCP servers so runs start grounded in the context your team already has.

    Integrations setup

A note on rollout

The safest adoption path is narrow: one team, one app workflow, one kind of request. Start answer-only, then grow into plans, previews, and pull requests as the review loop earns trust. Autonomy is opt-in throughout — nothing merges without your review.

Setup gets you a first run. These practical guides are how teams get great, repeatable value out of Gentlybot.

Writing great prompts

Name the outcome, hand over the context Gentlybot can’t infer, and scope the work — with before/after examples.

Choosing a mode

Ask, Plan, Implement, or Auto — what each does and how to climb the ladder as trust builds.

Reviewing and iterating

Read a plan, read the proof, and steer a thread with follow-ups all the way to a reviewed PR.

Automations & triggers

Schedule and event-driven runs — digests, CI fixes, bug-triage repros — that deliver as normal threads.

Playbooks by team

Concrete plays and example prompts for product, design, QA, engineering, GTM, support, and ops.