Developer agents for your SaaS product, an API for its AI features

Agents code in your repositories and open merge requests with a preview, and Hugo reviews them before you do. Your docs and Storybook stay current. For your own AI features, an OpenAI compatible API picks the model for you and shows every cost.

Create my spaceBook a demo
  • Your GitHub or GitLab repositories
  • A reviewed merge request before every merge
  • A visible cost for every job
Aurore Studio
Web app
Previewpreview-mr-214.your-saas.com
  • CSV export for invoicesDone
  • Fix the dashboard date filterIn review
  • Docs for the /invoices endpointRunning
Merge request ready to approveApprove

What holds a small product team back

A SaaS team has to ship features, fix bugs, write docs and answer customers, often with the same three or four people.

A growing backlog

Small fixes, customer requests, tech debt: tickets pile up behind the priorities.

Docs that lag behind

The docs and the Storybook describe the version from three months ago, and support pays for it.

AI costs that are hard to track

Several model providers, keys everywhere, and no clear view of what each feature costs.

What your agents take on

Developer agents work in your repositories, in an isolated session per task. Nothing reaches your main branch without review.

Work in your repositories

Nora on web interfaces, Julien on the API and data model, Marc on mobile. Each one does their part of the task, then hands over.

Reviewed merge requests

Every change arrives as a merge request, reviewed by Hugo, then by your team before merging.

Preview deployments

A preview for each change, on Vercel or Upsun, so you can test it before merging.

Docs and Storybook

An agent sets up the docs and the Storybook in the repository and keeps them in step with the code.

Chat that reads your docs

Your website chat reads your application's docs and links to the right page, or hands over to your support team.

AI API for your product

An OpenAI compatible API: your code calls a profile, ERHA picks the model and bills each call in euros. API keys per space, with restrictions.

How it gets set up

You connect your repositories, add the agents you need, and keep control over what gets merged.

  1. 01You connect your applicationsA project groups your applications (web, API, mobile, library), each with its GitHub, GitLab or ERHA hosted repository, and its Vercel or Upsun hosting.
  2. 02You add the agentsYou add the agents you need from the catalog: project lead, web interfaces, API, mobile, code review.
  3. 03You start tasksThe Start button splits the task between agents. It moves to review once everything is done, and your team merges.

An example: a bug reported in Slack

An eight-person SaaS startup collects customer feedback in a shared Slack channel.

  1. A message in the channel opens a task in the project, through a trigger set up by an admin.
  2. Julien fixes the API, Nora adjusts the screen, and a merge request opens with its preview.
  3. Hugo reviews the code, then a developer on the team tests the preview and merges.
  4. If the rule allows it, the agent's report is posted in the Slack thread, and the docs are updated in the same task.

Outcome

The team reviews and merges instead of writing every fix itself.

The ERHA building blocks involved

ProjectsYour applications, their repositories, previews, docs and Storybook.AgentsPatrice, Nora, Julien, Marc, Camille and Hugo, added from the catalog and set up for your space.APIAn OpenAI compatible API for your AI features, with model profiles, keys per space and spending caps.

Frequently asked questions

Can agents push code without review?

The delivery mode is set per application. With merge requests, nothing is merged without your team, and Hugo reviews the code first. When a decision belongs to a person, the agent asks and the task moves to blocked.

Is our code safe?

Agents work in an isolated environment, one session per task. No platform secret is passed to that environment, and spaces are isolated on every request.

Which models does the ERHA API offer?

OpenAI, Anthropic, Mistral AI, Google Gemini, DeepSeek, xAI Grok, Groq and Kimi. Your code calls a profile, not a provider: switching models does not require a code change.

How do we keep track of AI costs?

Every call and every agent job shows its cost in euros, computed by ERHA from its own price list. You set spending caps per space, and each API key can be restricted.

How is this different from a coding assistant in the editor?

A coding assistant helps the developer who is typing. An ERHA agent takes a task from the board, works on the repository in its own session, opens a merge request with a preview and reports back on the task.

Keep reading

TeamsTech and ITDevelopment, maintenance and documentation handed to agents on your repositories.SolutionsApp developmentWeb, API and mobile built by your agents, in your repositories.SolutionsAI website chatA chat on your site that answers from your documents and hands over to your team.IndustriesWeb agencies and studiosClient projects, developer agents in your repos, previews and cost per task.

Hand over your first piece of work this week

Open your space in a few minutes, or book a demo to scope a first use case with the ERHA team.

Create my spaceBook a demo