CLI Plugins
Konductro's CLI plugins bring the platform into the developer's IDE. They connect Claude Code, Amazon Q (Kiro CLI), OpenAI Codex, and Cursor to Konductro via MCP (Model Context Protocol), giving the AI in the editor access to project context, task details, and delivery workflows.
Why CLI Plugins Exist
Some workflows require access to the actual codebase — not an abstraction of it:
- Technical analysis needs to read real files, understand existing patterns, and trace dependencies
- Story decomposition needs to explore the codebase to scope tasks to specific files and repos
- Development benefits from having task context, acceptance criteria, and architectural notes available in the editor
These operations run on the developer's machine, where the code lives. The CLI plugin bridges the gap between the local codebase and Konductro's planning data.
Supported IDEs
| IDE | Plugin | Transport |
|---|---|---|
| Claude Code | @konductro/claude-plugin | MCP server (stdio) |
| Amazon Q (Kiro CLI) | @konductro/q-plugin | MCP server (stdio) |
| OpenAI Codex | @konductro/codex-plugin | MCP server (stdio) |
| Cursor | @konductro/cursor-plugin | MCP server (stdio) |
This page describes the published plugins: Claude Code 2.4.0, Amazon Q (Kiro CLI) 2.4.0, OpenAI Codex 2.4.0, Cursor 1.3.0. All four ship the same tools and commands.
The choice depends on which AI coding tool the developer uses.
What the Plugins Provide
Commands
Commands are guided, multi-step workflows that call the tools for you.
| Command | What it does | When you would use it |
|---|---|---|
start-ticket | Picks up an assigned ticket: confirms the branch, creates it, pushes the ticket spec and sets the ticket to In Progress. | When you begin work on a ticket. |
technical-analysis | Runs a technical analysis for a project phase against your local codebase and submits the document to Konductro. | When a technical analysis has been assigned to you. |
decompose | Breaks a story into implementation tasks by exploring the codebase, and submits each task with its developer brief. | When a story has been assigned to you for decomposition. |
regenerate-brief | Rewrites a task’s developer brief for your AI coding tool and the current code, and saves it as a new version. | When a brief is out of date, or written for a different tool than the one you use. |
bug-enrich | Investigates a bug against the codebase and adds structured technical findings to it. | When you are asked to triage a bug. |
prototype | Builds a UX prototype from the design system and approved documents, publishes it and submits it to Konductro. | When a prototype build has been assigned to you. |
find-prototype | Finds and downloads the approved UX prototype for the current project. | When you are building a feature that has a prototype. |
profile-repository | Scans the current repository and registers its profile with Konductro. | When a repository is new to Konductro, or has changed significantly. |
Starting a command in each AI coding tool
The four tools start a command differently. The table shows how, using the start-ticket command as the example.
| AI coding tool | How commands work | Example: start a ticket |
|---|---|---|
| Claude Code | Skills, started with a slash command. | /start-ticket |
| Amazon Q (Kiro CLI) | Ask in plain language. Kiro calls the Konductro tools to do it. | “What’s on my list?”, then “Start work on MRB-15” |
| OpenAI Codex | Skills. Ask for the workflow in plain language and Codex runs the matching skill. | “Start work on MRB-15” uses konductro-start-ticket |
| Cursor | Rules, applied when your request matches. Mention Konductro in the prompt. | “Start Konductro ticket MRB-15” applies the start-ticket rule |
Technical analysis and document templates
A workspace can define a Technical Analysis template under Settings → Documents, and a project can override it under the project's Settings → Documents. When a template applies, the document the plugin produces follows it: its sections, in its order, with any standing content reproduced as written. With no template at either level, the document uses Konductro's default structure.
The template is an instruction to the AI writing the document, not a check on the result. Review the document in Konductro as you would any other.
When a template applies, Konductro refuses a plugin that is too old to follow it: Claude Code, Amazon Q (Kiro CLI) or OpenAI Codex older than 2.4.0, or Cursor older than 1.3.0. The refusal says which version you need. Upgrading means installing the new version and re-running the plugin's setup, as described under Updating on the Installation page.
New to this? Your first technical analysis walks through the whole flow, from installing the plugin to finding the submitted document in Konductro.
MCP Tools
The plugins also expose individual MCP tools that the AI in the editor can call during conversation — listing tickets, loading context, submitting results, and more.
How It Works
The plugin runs as an MCP server using stdio transport. When the developer starts a conversation in their IDE, the AI can call Konductro tools to:
- Pull context — load task details, acceptance criteria, architectural notes, approved artefacts
- Execute workflows — start work, submit decomposition results, create PRs
- Push data back — submit technical analysis documents, repository profiles, prototype builds
All communication goes through the Konductro REST API using a personal CLI token. No direct database access, no secrets exposed to the IDE.
See Installation to get started, Your first technical analysis for a guided first run, or the Tools Reference for the full list of available tools.