Claude integrations for your CRM, documents and workflows
Claude connected to the systems you already run
Each build sits on a satori. platform and gets a fixed price before we start, so the number you approve is the number on the invoice.
Pick your builds
The Model Context Protocol (MCP) is an open standard, meaning an agreed way for AI assistants to talk to your internal systems, databases and tools with access rights you control.
What you get
- Connections to the CRM, the ERP and other business systems
- Controlled access to internal databases
- Claude that can create, update and look up records
- MCP servers built for your own systems
- Monitoring and running after go-live
Use cases
- Look up a customer record from inside a Claude chat
- Create tickets and tasks without switching tools
- Search internal documents and systems in one question
- Update records after Claude has analysed an email
Retrieval augmented generation (RAG) means Claude searches a copy of your manuals, policies and past answers before it replies, so it quotes your documents instead of guessing.
What you get
- Indexing of documents, PDFs and web pages
- Search by meaning across your knowledge base
- Answers based on your own documents
- The index updates when documents change
- A source reference in every answer
Use cases
- Internal knowledge search for staff
- Customer service answers based on your documentation
- Assistants for policies and regulations
- Onboarding help for new starters
AI flows that take over repeated steps, such as sorting incoming email or reading supplier invoices into your accounting system, and we say up front which steps a person still checks by hand.
What you get
- AI flows built for your own processes
- Connections to the tools you use today, such as Slack or Teams
- Triggers on events or on a schedule
- AI reading, sorting and summarising of data
- Monitoring and troubleshooting included
Use cases
- Sorting incoming email into the right queue
- Weekly reports and summaries written overnight
- Checking documents against a checklist
- Scoring and routing new leads
What it costs to keep a build running
The build itself gets a fixed price after the scoping call, and keeping it running afterwards has a published monthly price, all in euros excluding VAT.
satori-connect
€110 a month per integration
We fix the connection when an API changes or the link stops going through, and it is already included in satori-ai-department.
satori-run
From €230 a month per agent
We watch an agent we built or approved, fix it and update it when a vendor changes something, with a written operations report every month.
Frequently asked questions
MCP, the Model Context Protocol, lets Claude read from and write to your internal systems such as the CRM or the ERP. RAG, retrieval augmented generation, lets Claude search your own documents before it answers, and automation runs a repeated task on a trigger without anyone starting it. Most projects combine two of the three.
Yes, every build runs on satori-claude, satori-chatgpt or satori-mistral, because the connection needs the right licences, admin access and someone who answers when it stops working. Customers on satori-ai-department already have the platform included.
A single MCP connection or a RAG index over one document library typically takes two to four weeks, while an automation that touches several systems takes four to eight. We start every project with a scoping call and a written plan, which is where the fixed price and the timeline come from.
The build gets a fixed price after the scoping call, because two MCP connections are rarely the same size. Keeping it running costs €110 a month per integration with satori-connect, or from €230 a month per agent with satori-run, and satori-connect is already included in satori-ai-department.
Each connection only gets the access rights the task needs, read-only wherever possible, and you approve the list of systems and rights before we build. When data must stay in the EU, Claude runs on Amazon Bedrock with EU cross-region inference, meaning requests stay in EU regions, or on the EU multi-region endpoint of Google Cloud Vertex AI. Someone on your team still reviews that list whenever a new system is added.
Yes, and the combination is often where the hours come back: a RAG index lets Claude understand your documentation, and an MCP connection then lets it act on that knowledge in the ticket system. We scope the combination as one fixed-price project so nobody pays twice for the same groundwork.