Discovery call
One POM. We talk, scope the problem, decide if a paid block is a fit.
- No prep needed
- Honest go/no-go
- Email follow-up
Pay in POMs (1 POM = 25 minutes). Pay in agent-hours. Or pay the metered API rate, marked up to cover the wrapper.
Discovery is free. After that, pick the size of the problem. Every block is an outcome, not a clock.
One POM. We talk, scope the problem, decide if a paid block is a fit.
A focused block for one specific deliverable.
Half a day. Enough to ship something material end-to-end.
A full sprint week of senior AI-PM time. Ship the hard thing.
You bring the workflow, we run the agent. No infra, no on-call, no extra seat in your stack.
Always-on Claude agents running your operational workflows — research, monitoring, content, ops — billed by uptime, not seats.
Anthropic's list rates plus a flat $1 / MTok markup for monitoring, prompt-caching defaults, and one named human (Mat) for issues. Updated 2026-05-16.
| Model | Input ($/MTok) | Output ($/MTok) | Notes |
|---|---|---|---|
| Claude Opus 4.7 | 15 + 1 = $16 | 75 + 1 = $76 | Frontier reasoning + long-horizon agentic work. |
| Claude Sonnet 4.6 | 3 + 1 = $4 | 15 + 1 = $16 | Best balance of cost, latency, and capability. |
| Claude Haiku 4.5 | 1 + 1 = $2 | 5 + 1 = $6 | Fast and cheap, suited for high-volume agent tasks. |
Source of truth: claude.com/pricing#api. Cache-write, cache-read, and batch rates available on request.