mistral
claude
gpt
deepseekAI agents
Each agent has a role (developer, accountant, market watch). You write to it like a colleague and follow what it does.
- Role catalogue
- Messenger
- Your instructions
Hand over the work you no longer have time to do.
Sites and apps built, documents processed, customer requests handled. Agents take it on, you review before anything goes out, and every job shows its price in euros.
The models your agents use
- Anthropic
- OpenAI
- Gemini
- Mistral
- DeepSeek
- Grok
- Groq
- Kimi
Products
What can ERHA do for you?
Turn your know how into software
A repetitive or intricate task? A lapse in rigour or a moment of inattention can cost you dearly. Automate it: stop losing time and stop mistyping.
An agent keys in your delivery notes, compares your quotes, writes up your reports.
Keep your site and your apps current
A page to change, a price to correct, a feature to add? Ask for it, an agent prepares it and you approve it before it goes live.
Run your sites and your applications from your dashboard.
Give your team more hands
Accounting, sales, marketing, purchasing, quality, IT: an agent on call for every department, handling the daily work.
They work from your instructions, your documents and your tools.
Bring AI to the tools you have
Plug our agents into the tools you already use. Keep your habits and get the power of our technology.
Works with your mailbox, Slack, Teams, Drive or your ERP.
Five products in a single tool, with the same agents, the same colleagues and one invoice.
Projects and applications
Your ticket board, your applications, their code and their hosting. You move a ticket, an agent picks it up, and the change goes through a test version before it goes live.
- Source code
- Hosting
- Test version
Skills
Processing built for your documents. You drop a file and get back a report, a spreadsheet or a Word document, with its cost. On demand, or every Monday morning.
- Word, Excel, PDF
- At the time you choose
- Results and cost
Ask ERHA
Ask a question about your projects or tickets, it answers. It also helps you write a request.
- Questions about the space
- Getting started
LLM API
Your own software uses the same AI models as your agents, in the OpenAI format, with a meter per key.
- Compatible endpoint
- Restricted keys
- Usage per key
How it works
- 01
You describe the work
A ticket, a conversation, a file or a skill request.
- 02
An agent does it
It uses the model set for the task and shows you each step.
- 03
You approve
Nothing ships without your approval. The cost is shown.
Cost per job
Each run shows its price in euros.
Capped budget
A monthly budget per space, with an optional stop.
Isolated spaces
A space's data is visible only to its members.
Human approval
A blocked agent asks instead of guessing.
What you can hand over to your agents
Three kinds of projects, with the products they use.
Company website
A website to build or to keep improving.
- The agent codes the pages and opens a merge request.
- One test environment per change.
- Blog articles and cookie-free analytics.
Products used
- Projects
- AI agents
- Audience
- Blog
ERP and business application
An internal tool to build and maintain: ERP, CRM, extranet.
- Backlog, versions and planning in Tickets.
- Import tickets from Jira, GitHub or GitLab.
- Every delivery is reviewed and approved by you.
Products used
- Projects
- AI agents
Business skill
A repetitive process on your documents.
- IFS and BRCGS audit reports read and annotated.
- Laboratory reports written to Excel.
- Public tender monitoring and responses.
Products used
- Skills
- AI agents
- LLM API
You pay for what you use
No per-user licence. A space subscription and actual usage.
Space subscription
A monthly amount agreed with you.
Usage
Calculated by ERHA for each run, shown in euros.
€10 free
When you add your first payment method.
Frequently asked questions
How is this different from a ChatGPT subscription?
A chat answers you, it delivers nothing. In ERHA an agent has a job, your company history and access to your tools: it opens a merge request, produces an Excel workbook or redoes the same work every Monday, and you approve the result.
Do we need a technical team to start?
No. You describe the work in plain language, in a ticket or in a conversation. On sign-up the assistant creates your space, your first project and your first agent. For a business process, the ERHA team builds the skill and lets you know when it is available.
What does it cost, and how do we avoid a surprise at month end?
A space subscription agreed with you, plus actual usage: every run shows its cost in euros. You set a monthly cap beyond which runs stop, so nothing more can be invoiced. No per-user licence, and €10 free when you add your first payment method.
What happens when an agent gets it wrong?
Nothing goes live without you. The work arrives as a merge request with its test environment: you approve it, or you decline with a reason and the agent starts again. When information is missing, it asks instead of guessing.
Are our files used to train models?
No. Service data is hosted in France at Upsun, each space is isolated, and only the data a run needs reaches the model that performs it. When the contract ends, you export your files and they are deleted within 90 days.
Do we have to change our tools?
No. GitHub, GitLab (self-hosted included), Jira, Vercel and Upsun connect to your space, and tokens are encrypted. Your own software calls the same models through an OpenAI compatible API, metered per key.
Who chooses the AI model?
ERHA does, task by task, among Anthropic, OpenAI, Gemini, Mistral, DeepSeek, Grok, Groq and Kimi. You have no subscription or key to manage with each provider, and you get a single invoice.
Are we tied in for a fixed term?
No, the service has no minimum term. You cancel by email, it takes effect at the end of the current month, and the invoice only covers usage already incurred.
Give your agents a first job
Create your space, or show us one of your files in 30 minutes.