Skip to main content
This track is for technical builders, a developer or your AI Center of Excellence. What makes it distinct isn’t a single tool. It’s that you work in the coding harness your engineers already use, Codex, Claude Code, or the Gemini CLI, iterating at the code level to build compound agents. The non-technical track builds skills and simple agents. This track builds the compound ones.

​
The Agent Development Kit

The Agent Development Kit (ADK) is the throughline. It’s a disciplined sequence that runs from a business objective to a tested, deployed agent, grounded at every step in your Company Brain rather than guesswork.
There is nothing separate to install. The ADK methodology is served from the MCP itself. Once Klarity is connected in your harness, ask your assistant to build something and it calls get_agent_builder_instructions, then follows what comes back, fetching supporting material with get_agent_builder_resource as it needs it.
That has a practical consequence worth knowing. There is no plugin version to track and nothing to keep updated. The methodology your assistant follows is whatever the MCP is serving right now, so improvements reach you the next time you build.

​
The sequence

At every step the assistant can do the work, or your team can do it themselves. The ADK brings rigor and speed. The Company Brain is what unleashes it, because every step is grounded in observed work. The prompts below are what a builder types in their harness with Klarity connected.
1

Objective

Define the business objective and the technology stack the solution should live within. Giving both the goal and the stack constraints up front makes recommendations far more useful and deployable.
See Set the objective for what a well-formed one looks like.
2

Companion

Make sure the Company Brain covers the area in question. Where coverage is thin, deploy Companion against it before building.
3

Analyze

Produce a current-state read from observed data: pain points, bottlenecks, time and motion, systems in use, and bright spots.
4

Recommend and select

Generate candidate use cases and screen them down to the highest-leverage one that is feasible now.
Apply the screens in Pick the right use case. The Find transformation opportunities guide covers the tool chain behind this step.
5

Design

Produce the agent specification, effectively a PRD for the agent: scope, governance, credentials, and flow.
6

Build

Generate the agent: code, skills, and configuration.
7

Test

Generate and run evaluations. This is the hard but essential work of putting an agent through its paces before it ships.
8

Deploy

Ship into your environment.

​
What you build

Compound agents, managed, orchestrated, or custom SDK-built, and deterministic Integration Flows. See What you can build for how the primitives differ and when each one fits. Two disciplines run throughout. Every recommendation cites observed data, and a human signs off at each step.

​
Artifacts are part of the value

The current-state reads and recommendations come out as clear, visual HTML artifacts. The assistant generates them as it works. Showing these to non-technical stakeholders is one of the best ways to win business buy-in for a build.

​
Keeping agents current

An agent built against last quarter’s process drifts as the business changes. See Keep agents current for how to diff a repository of existing agents against the Company Brain and refresh what has moved.

​
Klarity support

Klarity’s FDEs can work alongside your engineers to build the agents, get them into production, and define the test and eval suites.

Agent Builder tools

The two tools that serve the methodology.

Build skills, agents, and automations

The same flow from any MCP client, not just a coding harness.