How we work.

Send us a message or book a 30-minute call. Tell us what you are trying to automate. We do not need a brief, a slide deck, or a detailed spec. A paragraph is enough to start. We respond within 24 hours on weekdays.

After one conversation, we send a 2-page scope document covering exact deliverables, timeline, pricing (setup fee plus monthly operating costs itemized: Claude tokens, voice minutes, telephony, vector storage, our retainer), human-in-the-loop checkpoints, and out-of-scope items. No scope document signed, no work started.

We build using Claude Code and Cursor on the Mastra or LangGraph orchestration stack with MCP integrations. You get access to a shared workspace where you can see progress. Standard build times: voice agent 5 to 7 days, WhatsApp support agent 7 to 10 days, RAG customer support 7 to 14 days, lead qualification 4 to 6 days, custom vertical agent 2 to 4 weeks. Langfuse observability is wired from day one.

We schedule a live demo with your team. You interact with the agent as a real user would. We document acceptance criteria before the session. If the agent does not pass, we iterate. We do not invoice the setup fee until you sign off. If we cannot deliver, you owe nothing.

On sign-off, you receive: full codebase in a private GitHub repo, CLAUDE.md documenting the architecture, .cursor/rules/ files for your team, MCP server configs, a runbook covering prompts, intents, and failure handling, and Langfuse dashboard access. We train your team on the codebase if needed.

Agents are not set-and-forget. Models update, prompt patterns drift, and buyer behavior shifts. The optional retainer covers monthly prompt audits, token budget monitoring, model updates, up to 5 hours of changes per month, and priority 4-hour response time. Retainer pricing: $499 to $8,000 per month depending on agent complexity. We recommend it.