Use cases › Software development

A development coworker in the team’s Slack channel

Real example: an agent we build and run ourselves, every day.

In short. An Android development team hands the agent implementation work, investigations and code checks, in a Slack channel everyone can see. The agent proposes changes by opening a pull request on GitHub; the team reviews it, tests it and decides whether to accept it.

The context

In a development team, a lot of time goes into work that is necessary but repetitive: tracking down where a bug comes from, working out how a part of the code behaves, making well-defined changes. Many AI coding tools are tied to an individual developer: the context stays on one person’s computer or account, and what the assistant has “learned” about the project isn’t shared with the rest of the team. When that person is on holiday or moves to another project, the context is lost.

What the agent does, step by step

Gets the request in the channel.

Someone on the team posts a task, a bug or a question in the shared Slack channel.

Rebuilds the context.

It reads the relevant code and uses what it remembers about the project: structure, conventions, past decisions.

Investigates or implements.

For a bug, it looks for the cause and explains it; for a well-defined task, it prepares the change, following the team’s conventions.

Uses the project’s integrations.

For example, it reads a design from Figma, takes Greptile’s code analysis into account and checks the project knowledge organised in Open Knowledge Format (OKF).

Opens a pull request on GitHub

with the change and an explanation of what it did and what to check.

Reports back in the thread

with the pull request link, open points and questions, if information is missing.

Updates the project memory

with what the team decided, so the next request starts from there.

How the team works with it

Channel

A shared Slack channel. Anyone on the team can post a request, from their phone too. Everyone sees the request, the progress and the result, and anyone can step into the thread to adjust the direction.

Who approves what

What the agent can do on its own and what needs approval is agreed at the start. Code never enters the project on its own: every change goes through a pull request that a team member reviews, tests manually and decides whether to accept.

Who changes the rules

The agent’s instructions and rules are updated with us, when the project or the way of working changes.

Benefits

The team stays in control. Every change goes through a pull request that a person reviews, tests and accepts.
Memory shared across developers. The project context is available to the whole team, not just one person.
Visible work. Requests, answers and feedback live in the channel: nobody works in the dark.
The history stays with the company. Memory and activity history are kept in the company’s dedicated environment.
Built around the project. Skills and integrations built on the tools the team actually uses.
No laptop needs to be on. The agent works in its own environment, not on someone’s computer.

What the fee covers and what’s on quote

Included in the Rilvo Coworker fee

€249/month + VAT

Activation, the Slack channel, instructions and rules for your project, repository access with the agreed permissions, GitHub pull requests, shared memory, AI usage subject to reasonable use, maintenance and human support.

Separate, on quote

Custom skills and integrations, such as those with design tools (Figma), code analysis tools (Greptile) or knowledge management (OKF). You receive the source code of what we build for you.

See pricing → · Related setup: Rilvo Coworker for software teams →

FAQ

Can the agent change the code on its own?

No. It proposes changes through a pull request on GitHub. A team member reviews it, tests it and decides whether to accept it.

Does our code leave the dedicated environment?

Memory and history stay in the company’s dedicated environment. To do its work, though, the agent sends parts of the code to the AI model provider, and pull requests live on GitHub. If you use external services such as Greptile, they analyse the code too. We go through this together at the start, with the minimum permissions needed.

Does it only work for Android?

No. The example comes from an Android project, but the same pattern works for other software projects: a shared channel, project memory, pull requests reviewed by the team.

Does every developer need an account or a licence?

No. The agent works in the team’s Slack channel and the fee is per company, not per person.

More use cases

Internal project

Every week, a website article built from the Facebook posts

An automatic job prepares the WordPress draft; a person approves and publishes it.

See how it works →

Example scenario

Know-how stays with the company

The team’s procedures and answers in one shared memory, even when people change. The team decides what gets saved.