Google Releases Conductor: a context-driven Gemini CLI extension that stores information as Markdown and organizes agent workflows

Google has launched Conductor, an open source preview extension for the Gemini CLI that turns AI code generation into a streamlined, context-driven workflow. Conductor stores product information, technical decisions, and work plans as a Markdown version within the repository, then calls Gemini agents from those files instead of ad hoc chat messages.

From conversation-based coding to context-driven development

Most AI code today is session based. You paste the code into the dialog, define the function, and the context disappears when the time runs out. The operator takes that as a core problem.

Instead of ephemeral notifications, Conductor maintains a persistent context index within the repo. It captures product goals, limitations, technology stack, workflow rules, and style guides like Markdown. Gemini then reads these files on every run. This makes AI behavior replicable across machines, shells, and team members.

Conductor also uses a simple lifecycle:

Context → Specification and Application → Use

An extension does not jump directly from a natural language application to coding. It first creates the track, writes the spec, produces the program, and then just does it.

Installing Conductor in the Gemini CLI

Conductor works as a Gemini CLI extension. To enter a single command:

gemini extensions install  --auto-update

I --auto-update tag is optional and keeps the extension in sync with the latest release. After installation, Conductor commands are available from within the Gemini CLI in the project directory.

Project setup with /conductor:setup

The workflow starts by setting the project level:

This command uses an interactive session that creates a base context. Conductor asks about the product, users, requirements, technology stack, and development processes. In these responses it produces a conductor/ a directory with several files, For example:

• conductor/product.md
• conductor/product-guidelines.md
• conductor/tech-stack.md
• conductor/workflow.md
• conductor/code_styleguides/
• conductor/tracks.md

These artifacts define how the AI ​​should think about the project. They define target users, high-level features, acceptable technologies, testing expectations, and coding principles. They live in Git with all the source code, so changes in the context are updated and readable.

Tracks: specify and organize as first-class artifacts

The conductor introduced tracks representing units of work such as features or bug fixes. You create a track with:

or in short description:

/conductor:newTrack "Add dark mode toggle to settings page"

For each new track, the Conductor creates a subdirectory conductor/tracks// contains:

• spec.md
• plan.md
• metadata.json

spec.md carries detailed requirements and track restrictions. plan.md contains a step-by-step implementation plan divided into phases, tasks, and sub-tasks. metadata.json stores identifiers and status information.

Conductor helps to specify the draft and use existing context files. The developer then edits and approves them. The important point is that every implementation must follow a clear plan and version control.

Implementation with /conductor:implement

When the program is ready, manual control on the agent:

Conductor is studying plan.mdit selects the next pending task, and uses the default workflow. Common cycles include:

1. Check the corresponding files and context.
2. Suggest code changes.
3. Do exercises or test by saying conductor/workflow.md.
4. Update job status at plan.md and the rest of the world tracks.md.

The extension also adds checkpoints to section boundaries. At these points the Conductor pauses to reassure the person before continuing. This prevents the agent from using large, unupdated objects.

Several operating instructions support this flow:

• /conductor:status shows the track and progress of the work.
• /conductor:review helps ensure work completed against product and style guidelines.
• /conductor:revert uses Git to roll back a track, section, or task.

Returns are defined in terms of tracks, not hashes of raw commits, which is easy to think of in most work changes.

Brownfield projects and team work flow

Conductor is designed to work on brownfield codebases, not just new projects. If you run /conductor:setup in an existing repository, the context session becomes a way to extract the abstract information from the group into clear Markdown. Over time, as more tracks become active, the content directory becomes a unified representation of the system’s architecture and limitations.

Group-level behavior is coded workflow.md, tech-stack.mdand style guide files. Any developer or AI agent using Conductor in that repo gets the same rules. This is useful for enforcing testing strategies, aggregating expectations, or approved frameworks for all contributors.

Because the code and programs are in Git, the code can be updated, discussed, and changed in the same way as source files.

Key Takeaways

• Conductor is a Gemini CLI extension for context-driven development: It’s an open-source, Apache 2.0-licensed extension that works within the Gemini CLI and calls AI agents from local Markdown context instead of ad hoc notification.
• The project content is saved as a Markdown version underneath conductor/: Similar files product.md, tech-stack.md, workflow.mdand code style guidelines define product goals, technology choices, and workflow rules that the agent reads on each run.
• Work is organized into tracks by spec.md again plan.md: /conductor:newTrack creates a directory containing the track spec.md, plan.mdagain metadata.jsonmaking requirements and implementation plans transparent, reviewable, and integrated with Git.
• Implementation is controlled by /conductor:implement and ops have no clue: The agent performs tasks according to plan.mdupdates continue on tracks.mdand foundations /conductor:status, /conductor:reviewagain /conductor:revert for progress checks and Git-backed commits.

Check it out Repo and Technical Details. Also, feel free to follow us Twitter and don’t forget to join our 100k+ ML SubReddit and Subscribe to Our newspaper. Wait! are you on telegram? now you can join us on telegram too.