Project — Sprint / SKU 02

MCP / Agent Integration Sprint

Ship one MCP server or agent integration end-to-end in 2 weeks — with evaluations and a handoff doc your team can maintain.

Fixed fee

$20k

Duration

2 weeks

Engagement

Standalone — no retainer required

Start onboarding →Book a 15-minute fit call ← All sprint SKUs

The problem this solves

Why this exists

MCP integrations look easy in a demo and hairier in production. Auth, schema design, error handling, evals, observability, retries — you can spend a month rediscovering all of them.

We have shipped these before. You get the version that works on day 14, in your stack, with documentation your next engineer can ramp from.

Is this for you?

Concrete fit and anti-fit

If we're not a fit, we'd rather you find out here than after a sales call.

Good fit

▹You want to add an MCP server to an existing agent surface (Claude Desktop, Cursor, your own product) — to expose your data, your tools, your domain.
▹You have a known integration target and an authoritative source of truth for the data behind it.
▹Your team can maintain it after handoff but does not have spare capacity to build it now.
▹You want it shipped, evaluated, and documented — not a half-built proof of concept.

Not a fit

×You don't know what MCP is and don't have an integration target — that is a different conversation; book a discovery call.
×You want a 'platform' for many integrations — that is a multi-sprint or Retainer engagement.
×You need it shipped in less than a week — schedule honestly; this is not an emergency unit.
×You expect us to fix your underlying data source before exposing it — that is a separate scope.

Schedule

Week by week

What happens, in order, with checkpoints. The SOW dates match this schedule.

Week 1
1. Scope and build
- ▹60-minute kickoff. Lock target client (Claude / Cursor / your own), tools and resources surface, auth model, data shape.
- ▹Build the MCP server skeleton in your stack (TypeScript by default; Python on request).
- ▹Implement core tools and resources, wire to your data source.
- ▹Mid-sprint demo so you see it working before week 2.
Week 2
2. Harden and ship
- ▹Evaluation harness with test cases your team can extend.
- ▹Error handling and observability hooks (logs and metrics in your stack).
- ▹Handoff documentation: README, runbook, how-to-extend guide.
- ▹Final demo and 60-minute handoff call.
- ▹2 weeks of async support kicks off after final delivery.

Deliverables

What you have at the end

Each deliverable has a format. We don't ship slide decks instead of code.

01

MCP server

Production-ready, deployable. We ship the code; you deploy unless a deploy add-on is scoped.

FormatSource in your repo, your stack

02

Evaluation harness

Test cases your team can extend. CI-ready; we wire one example pipeline if your CI is straightforward.

FormatTest suite in your repo

03

Observability hooks

Structured logs and key counters. We use what you already pay for (Datadog, Honeycomb, Grafana, Sentry, etc.).

FormatLogs and metrics in your stack

04

Handoff documentation

README, runbook, and a how-to-extend guide written for the next engineer who has never seen the code.

FormatMarkdown in your repo

05

2 weeks of async support

Bug fixes, clarifications, and small adjustments after final delivery. Not a maintenance contract — that is a Retainer.

FormatEmail / Slack, included

What you provide

Concrete dependencies, not 'share relevant info'

If these aren't in place, the ship date moves. We say so up front.

▹A named decision-maker for acceptance and scope changes (~30 minutes/week).
▹A named technical contact for access and clarifications (~2 hours/week).
▹Read or write access to the data source the MCP server will expose, with least-privilege credentials.
▹Decision on auth strategy at kickoff (API key, OAuth, mTLS, etc.).
▹Decision on deployment target at kickoff (you deploy, or scope a deploy add-on).
▹Written responses to our questions within 2 business days.

Out of scope

What this does NOT include

Read this. If something you assumed is here isn't, raise it on the fit call before we sign.

×Marketing the integration to MCP directories or app stores (we can do it as an add-on).
×Long-term maintenance after the 2-week support window — that is a Retainer.
×Building a custom MCP client (we ship MCP servers; client work is separate).
×Integration with PHI sources without an executed BAA.
×Cleaning up an upstream data source that is not query-shaped (separate sprint).
×Any work not expressly listed in the weekly schedule or deliverables above.

Success looks like

One paragraph, plain English

Two weeks from signing, your data is live in your agent surface, the eval harness is green, and your team can extend it without needing to ask us first. The handoff doc is the durable artifact — your next engineer ramps from it.

What happens after

Common post-sprint paths

What clients typically do after this sprint ships. None of these are required.

▹Async support window for 2 weeks (bug fixes, small clarifications).
▹About 50% convert to a Retainer for ongoing maintenance and extensions.
▹Some clients run the same SKU again to ship a second integration once the first proves out.

Pricing and terms

What it costs and when you pay

The same numbers appear in the SOW. No bait-and-switch on the legal.

Fixed fee: $20,000 USD, fixed.
Payment schedule: 50% on signing ($10,000), 50% on final delivery ($10,000).
Net terms: Net-15 from invoice date.
Cancellation: Cancel for convenience: you owe work performed plus 25% of the unbilled remaining fixed fee. Cause-based termination follows the MSA.
Founding-client discount: Up to 20% off, first 3 months only, capped to the first 3 clients to sign at any tier.

FAQ