- Home
- AI development
- Claude development
Claude development
Claude set up for your staff, your apps and your software work
Claude is the family of AI models from Anthropic. We put it to work four ways: as a tool for your staff, as the model inside an app we build for you, as Claude Code agents that do software work, and connected to your own systems. Which one fits depends on who will use it and where the work happens.
- Prices published on the site
- First working version in two weeks
- Support month to month, cancel any time
Claude is the model, and you choose where it works
Anthropic's Claude can be reached in several ways: in its own chat apps, through an API that software calls, in a coding tool called Claude Code, and through connectors to other systems. Each suits a different job and bills differently.
We do not assume Claude is the best model for your task. On your own 20 to 50 cases we test it against Gemini or an open-weight model, then use whichever does better at a price you accept. The method is on the evaluation page.
Four ways to put Claude to work
| Claude for your staff (Team or Enterprise plan) | Claude inside your own app (API) | Claude Code agents (software work) | Claude connected to your systems (MCP) | |
|---|---|---|---|---|
| What it is | Claude's chat apps for the whole team, with admin controls and shared projects | Claude called by software we build, with your rules, checks and logging around it | Claude reading a code repository, running commands and editing files for you to review | Claude, in the apps above, given narrow access to your database, CRM or files |
| Who uses it | Office staff, managers, anyone who writes, reads or analyzes all day | Your staff or customers, through a screen or form we build; they never see Claude | Developers, and anyone who starts a task from a board | The same people, asking questions in plain words about live data |
| What we build | The workspace in your company's name, shared instructions for your common tasks, a short session for your team | The app or workflow: prompts, checks, an approval screen, a log of every call | The agent setup: instructions, permissions, a test on a real ticket. See Cardloop | The MCP server: a few narrow tools, sign-in per user, an audit log |
| How Anthropic bills it | Per user, on a plan | By usage, on your own account | On a plan seat or by usage, depending on the setup | Through the plan or API usage of whoever is asking |
Prices for plans and usage are on Anthropic's own pages and change, so we check them with you at the time you sign up.
Four jobs we build on Claude
Read and extract
Reads PDFs, emails and forms inside your app and returns the fields in a fixed layout, with a flag where it is unsure.
Draft and summarize
Writes replies, reports and summaries from your records for a person to approve. See generative AI for documents.
Act in your systems
Looks things up and prepares the next step in your CRM or accounting system, with approval gates. See AI agents.
Run software work from a board
Cardloop, our open-source (MIT) cockpit, runs Claude or Codex agents from a kanban board, one git worktree per task, with a live browser pane. We run our own projects on it.
When to build on Claude, and when to skip the build
A good fit when
- Your team writes, reads or analyzes text all day and one shared tool would save hours
- You want AI inside your own app or process, with your rules, logging and approvals
- You maintain software and have a backlog of small, well-defined tickets
Something simpler will do when
- You need one answer once: use a chat app yourself, no build needed
- A fixed rule does the job: a script or workflow tool costs less and never varies
- Another model tests better on your cases: we say so and build on that one
Staff setup by the hour, apps and agents by the project
One workflow running end to end, or one internal tool.
Several systems made to work as one: integrations, roles, reports.
Custom software built from scratch, over a longer schedule.
Setting up Claude for your staff is small enough to bill hourly, $150 to $200 an hour. An app or agent on one task usually lands in one process, and one that touches several systems is connected systems. Anthropic bills plan seats or API usage to your own account, and we estimate the usage in the audit.
Care $300 to $500 a month. Active $750 to $1,500 a month. AI model usage, GPU time and hardware are paid to the provider, on accounts in your name.
Questions about building on Claude
Do we need the API, or is a Team plan enough?
A Team or Enterprise plan is enough when your people use Claude themselves in a chat window. You need the API when Claude has to work inside your own software: reading incoming PDFs into your system or filling in a CRM. Some scheduled work can run in Claude's own apps.
Is Claude the best model for our work?
We cannot say until we test it. We run Claude, Gemini and, where it makes sense, an open-weight model such as Llama or Qwen on 20 to 50 of your own cases and compare the results and the cost. See LLM evaluation.
Does Anthropic train on our data?
By default Anthropic does not train on inputs to its commercial products, which include Team, Enterprise and the API (provider terms as of September 2026; see also private AI over your documents). Personal accounts follow different terms, so staff should not paste client files into them. Before we build, we write down what data the system sends.
What is Claude Code, and do we need it?
Claude Code is Anthropic's coding agent: it reads a repository, runs commands and edits files. It is useful if you maintain software and have well-defined tickets to work through. If you have no codebase you do not need it. We run our own projects through Cardloop, which starts Claude or Codex agents from a kanban board.
Can Claude connect to our database or CRM?
Yes, through an MCP server that we build. It lists a few narrow tools, checks who is asking, runs read-only unless a write is approved, and logs every call. See MCP servers.
Who owns the accounts and the code?
The Anthropic account is opened in your company's name and billed directly to you, and the code lives in your repository; the custom code we write is assigned to you under the contract once paid for. Support is month to month, and you can take the system to another team at any time.
Related services
- AI agents Agents that finish a task, with your sign-off
- MCP servers Connect AI assistants to your own systems
- LLM evaluation Test which model does your job best
- Generative AI Drafts, summaries and documents from your data
- Cardloop Our open-source cockpit for AI agents
- Private AI over your files Answers from your own files, on your server
Tell us what's slowing you down.
You get a free written review by email within two business days.