Brief
You tell us what you need. Use our brief form — no technical knowledge required. Just describe the problem, the process, and what a good outcome looks like.
How we work. Five steps to a working agent.
NDA and discovery
Before you share anything sensitive, we sign an NDA. Then we get into the detail — your current process, your tools and systems, your data, your team. We map how things work today and identify exactly where the agent sits in that flow. This stage is often where we find the real opportunity, which isn't always where you expected it.
Scoping and agreement
We come back with a clear scope — what we'll build, what it will do, what you'll get, and what it costs. No hidden extras.
Build
We build the agent. Fully structured — system prompt architecture, tool integrations, MCP server connections where needed, skills and memory configuration, testing against real scenarios.
Handover
You get the working agent, full documentation, and a 30-day support window. All IP is entirely yours.
What you need to provide
- Description of the problem and current process
- Access to relevant context (documents, data structures, example inputs/outputs)
- 1–2 hours for discovery
- Feedback during the build
What you don't need
- Any technical knowledge
- Your own AI tools or API keys
- A fully formed spec
Common questions
- How long does a build take?
- A single agent typically takes 1–2 weeks. Multi-agent builds take 3–6 weeks.
- What AI models do you use?
- Primarily Claude (Anthropic). We advise on the right model for your use case.
- Do I need to host anything?
- We advise on options. Many agents run via API with no hosting required.
- What happens after the 30-day support window?
- We offer ongoing maintenance packages. Models and tools evolve — keeping agents updated is a real need.
- Is my data safe?
- We sign an NDA before you share anything. We don't retain your data after delivery.