Features

Built for the messy apps real teams maintain.

Gently sets up your project and gives each thread an isolated environment for the real app: frontends, APIs, background jobs, and backing services. Your team can work in parallel without managing package installs or competing over one shared development environment.

Multiple apps in one working environment
Real routing across paths and subdomains
Backing services started and checked together

Gently handles the environment

Gently gets your project running with minimal supervision: installing packages, resolving version conflicts, starting services, and checking the app. PMs, designers, and the rest of the team can get to work without following an install guide.

Each thread has its own isolated development environment. If an experiment breaks one, other threads keep working. The whole team does not have to wait for someone to repair a shared environment.

Let Gently set up your project

One connected project

Crosstown’s repositories and saved setup

  1. Thread 1 · Product

    Explore delivery options

    Own code, dependencies, services, and preview

  2. Thread 2 · Design

    Sketch courier availability

    Own code, dependencies, services, and preview

  3. Thread 3 · Engineering

    Build recipient tracking

    Own code, dependencies, services, and preview

Work together inside a thread. Explore independently across threads. Gently can diagnose and fix setup problems in the conversation.

Acme commerce platform
Simplified architecture
9 services live
Public traffic
shop.acme.testadmin.acme.testapi.acme.test/v1
Edge router
Routing service
Matches each hostname and path to the right app
Application layer
Storefront
Frontend
Nuxt · port 3000
shop.acme.test
Admin console
Frontend
React · port 5173
admin.acme.test
Commerce API
Backend
Rails · port 8080
api.* /v1
Identity service
Backend
Node · port 4000
internal only
Job worker
Worker
Background process
queue consumer
Shared infrastructure
Postgres
Database
Primary data
Redis
Cache + queue
Jobs and sessions
Object storage
Storage
Uploads and assets
One sandbox, the whole system. Gently starts every node, wires its dependencies, publishes its routes, and checks the complete architecture together.

Engineering asks

Couriers need to set when they can work, so operations can offer each route to someone available that day.

Task prompt

Let couriers set their availability, and only offer routes to couriers available that day.

Result: Couriers choose their available days, and route offers respect that schedule. The day toggle persisted after a browser reload; all 64 API tests passed after review fixes.
Explore the environment, developer tools, reviews, and pull requests →

Ask about product behavior, implementation details, customer-facing copy, or release history and get a cited answer from real sources.

  • Product Q&A
  • File and PR citations
  • Questions and follow-ups from Slack
  • Plain-language summaries

Gently handles setup, package installs, tool versions, and services. Each thread gets its own environment so PMs, designers, QA, and engineers can work in parallel without maintaining local development setups.

  • Agent-led project setup
  • Isolated thread environments
  • Full-app preview URLs
  • Setup fixes in the conversation

Keep prompts, decisions, screenshots, recordings, checks, and code review notes in one thread so every function can contribute without bypassing engineering review.

  • Thread history
  • Screenshots and videos
  • CI feedback
  • Reviewer comments

Gently in Slack

Support needs an explanation. Sales needs to check a promise. Product needs to understand a flow. Ask Gently where your team already works, with answers grounded in your connected code and tools.

  • Mention Gentlybot or send a DM. Describe the question and attach a screenshot when it helps.
  • Keep the conversation in Slack. Read the answer, ask a follow-up, and answer Gently’s questions in the same thread.
  • Open the work in Gently. Follow the thread’s link when you want to explore previews, wireframes, or code changes.
How to ask from Slack

Example Slack exchanges · Crosstown, a fictional delivery app

# customer-support
M
Maya10:42 AM

@gentlybot In Crosstown, what happens when a courier cannot finish a delivery?

GentlybotAPP10:43 AM

The courier marks the stop “Could not deliver” and chooses a reason. If the customer has an email address, Crosstown tells them the delivery failed and the order is returning to the shop. The shop arranges another delivery.

In a channel, mention Gentlybot again in your thread reply to ask a follow-up. The conversation continues in the same Gently thread.

Connect Slack →
Wireframes

Ask Gently to sketch a few ways a change could work. Put the rough options in front of teammates or customers, find out what is clear or confusing, and revise them while changes are still cheap. Choose a direction before spending time on detailed design and implementation.

  • Compare different approaches to the same problem
  • Keep feedback focused on the flow and the choices
  • Refine the sketch in the thread before deciding what to build

151 threads produced wireframes · Production usage across our first teams, May–August 2026.

Product and design asks

Couriers need a simple way to choose the days they can work. The team wants to compare two options.

Task prompt

Sketch two simple ways for couriers to set which days they are available to work.

Read the full task prompt

Sketch two simple ways for couriers to set which days they are available to work. Follow-up: Show both as compact, rough sketches side by side so we can compare them in one view. Follow-up: Keep the two options side by side in a 650px-wide thread, with the whole sketch under 450px tall. Follow-up: The sketch clips the bottom controls. Make both cards fit fully without internal scrolling.

Result: Gently sketches a recurring weekly schedule and a date-by-date option for comparing how couriers choose their working days.

Option A · Weekly schedule

Choose the weekdays a courier usually works. Ask whether a recurring schedule fits their routine.

Option B · Upcoming dates

Set availability for specific dates. Ask whether this makes changing plans and taking days off easier.

Product management asks

A new teammate wants to understand how an order gets from a shop to its customer.

Task prompt

Show me a simple diagram of how an order gets from a shop to the customer in Crosstown.

Read the full task prompt

Show me a simple diagram of how an order gets from a shop to the customer in Crosstown. Follow-up: Make the diagram compact: six main steps in two rows, so it is readable at a glance.

Result: A diagram follows the order from upload and route planning through courier acceptance, delivery, and the customer notification.
Diagrams

Ask how a request flows, what happens when a user clicks, or how two services talk, and the answer arrives with a diagram traced from the real implementation — not a wiki page that went stale last quarter. Teams use them to onboard new hires, untangle UI flows, and weigh architecture changes with the trade-offs visible.

  • “Diagram how requests reach the worker queue” — one prompt
  • UI flows traced through the real event handlers
  • Research and architecture answers with the trade-offs drawn out

64 threads asked for a diagram · Production usage across our first teams, May–August 2026.

Proof, not promises

Passing tests aren’t the same as a working product. When a change warrants it, Gently opens the running app on its live sandbox URL, walks the changed flow like a user would, and looks at the screen with vision models to confirm the change behaves. The screenshots and recordings land in the thread and on the pull request — so review means watching the change work, not trusting a diff.

  • A real browser, not a guess
  • Evidence where reviewers look

745 threads checked their own work in a browser — most looked more than once

QA asks

A courier accepts a route, delivers one order, and cannot deliver another. QA needs to check what happens next.

Task prompt

Test that a courier can accept a route, deliver an order, and mark another delivery as failed. Check the customer emails too.

Read the full task prompt

Test that a courier can accept a route, deliver an order, and mark another delivery as failed. Check the customer emails too. Follow-up: Re-test accepting a route, delivering one order, and marking another delivery failed in the browser. Use demo orders with customer email addresses and check both emails. Follow-up: Summarize the completed browser checks and the two verified emails in a short checklist.

Result: Gently completed the courier flow in the browser and verified both the delivery confirmation and failed-delivery email in Mailpit.
Automations

Save a prompt, then run it on a schedule or behind a connected event from GitHub, Slack, Jira, Linear, or Notion. Every automation delivers as a normal thread — same sandbox, same evidence, same review trail — so recurring work gets done without becoming unreviewed work.

Schedule trigger
Every weekday at 7:00

Review yesterday’s merges for risky changes and post a digest to #engineering.

GitHub trigger
CI fails on an open PR

Reproduce the failure in a sandbox, fix it, and push to the same branch.

Slack trigger
New report lands in #bug-triage

Try to reproduce it, find the likely cause, and reply in the Slack thread with findings.

GitHub trigger
Dependabot opens a PR

Boot the app in a sandbox, click through the affected flows, and flag anything broken.

Linear trigger
A customer-impact issue is created

Check the report against the product and add likely causes plus the next diagnostic step.

The operating model

Teams can adopt Gently gradually: answer-only first, then planning, full-app previews, testing, and pull requests as trust grows.

86% of PRs opened from threads get merged · median 1.2 h from prompt to merge

01

Ask

Anyone on the product development team can ask how the app works, what changed, or whether a customer request is possible.

02

Plan

Gently turns fuzzy requests into a reviewable plan with scope, risks, affected app surfaces, and test ideas.

03

Preview

Approved changes run inside the app sandbox with a live URL — plus screenshots and browser recordings when the change calls for them.

04

Review

The final output is a pull request with context, checks, and a clear trail for engineering to approve before production.

Collaborative by default

Gently keeps the product question, files, decisions, preview, and review trail together, so each role can pick up where the last person left off instead of restarting the conversation.

Product frames the work

PMs can turn customer feedback, roadmap notes, or Slack context into a clear product direction.

Design shapes the experience

Designers can refine flows, copy, states, and visual direction while keeping the same product context.

Engineering reviews and builds

Engineers can continue from the same thread, review the evidence, and send feedback back through the loop.

Gentlybot Desktop

Turn your Mac into a Gentlybot runtime for local execution, live HTTPS previews, and local AI endpoints. Keep the same shared thread and review workflow while the app runs on your machine.

Explore Gentlybot Desktop
Local runtime
macOS · Apple Silicon
  • Run app sandboxes with Docker on your machine
  • Open local apps through trusted HTTPS previews
  • Connect Ollama, LM Studio, or another local AI endpoint

Gently can work across OpenAI, Anthropic, and open-source model families. Your team gets capable models out of the box, and Enterprise customers can bring their own model accounts or private deployments.

Provider resilience

Gently threads retain their product context across turns. When a model provider has an outage or rate-limit issue, a session can be continued with another provider instead of starting over.

  • Keep thread history, files, and decisions in context
  • Move between supported providers when needed
What makes it different

Regular AI tools

Help with one-off apps, toy demos, simple repos, or generic advice, but struggle when real products span multiple frontends, backends, workers, permissions, routing layers, and deployment rules.

Gentlybot

Works with the real shape of your app: connected repos, services, workers, routes, permissions, sandboxes, evidence, and reviewable artifacts your team can trust.

Not every team needs to implement. Some people only ask, some plan, some test, and some approve builds. Gently can match that rollout across the whole product development team.

Request access and tell us what makes your app complex. We will help you start with a safe, reviewable setup for the team members who need it first.

Request access