---
title: "Run an agent on your repository"
type: HOW-TO
updated: 2026-09-23
---

# Run an agent on your repository

Install an agent from the Marketplace, start it on your repository, and review its work before approving a merge.

Last updated 2026-09-23

## Install and start

1. Open your repository and the Marketplace.
2. Choose an agent from the [supported agents](/docs/agents).
3. Review its requested permissions and install it.
4. Start it on your repository and complete its vendor sign-in if requested.
5. Give it the task and inspect its progress.

## Use separate workspaces

Each agent works in its own isolated workspace. Several agents can run on the same repository at once. Review and merge each result separately; isolation does not remove conflicts when their changes meet.

## Architecture access

Every agent receives Panolayer's tools for targeted lookup of symbols and relationships in Architecture. They can request the context they need; they are not given a full architecture dump.

## Review the result

1. Open the completed run and inspect its changed files.
2. Read the verification results and investigate failures.
3. Decide whether to approve a merge into your branch.

An agent can’t merge its own work into your branch without your approval. You can review the results and merge with failures.

## Execution and billing

The twenty curated third-party agents run their command-line interfaces inside hardened Linux containers. They use the vendor’s own account, subscription, or API-key options. Panolayer does not pay their model bills. Panolayer Agent uses Panolayer credits included in your plan.

## Continue learning

- [Supported agents](/docs/agents)

- [How agent work reaches your branch](/docs/concepts/how-agent-work-reaches-your-branch)

- [Review changes and commit](/docs/changes)