> ## Documentation Index
> Fetch the complete documentation index at: https://docs.runlayer.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Create and update

> Publish artifacts with Runlayer Agents or a connected MCP client, then revise them at the same link.

## Create with a Runlayer Agent

Ask the agent for a file and explicitly ask it to publish a shareable link. Include
the format, audience, and information the artifact should contain.

<Steps>
  <Step title="Describe the result">
    For example: "Build a self-contained HTML dashboard from this CSV. Include
    filters and a summary chart. Publish it as an artifact and give me the link."
  </Step>

  <Step title="Review the published file">
    Open the link in the agent's response. You can also find the file in the
    agent's **Artifacts** tab or the main **Artifacts** library.
  </Step>

  <Step title="Revise and share">
    Ask for changes to the existing artifact, then review its
    [sharing settings](/artifacts/sharing) before sending the link to colleagues.
  </Step>
</Steps>

Artifact publishing is enabled by default for new agents. If publishing is
unavailable, ask the agent owner to check the agent's settings.

### How agent publishing works

The agent creates a file in its workspace and uses these built-in tools:

| Tool                  | Purpose                                                                                                                                                            |
| --------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| `publish_artifact`    | Publishes a file under `/tmp/workspace` and returns its artifact ID and stable URL. Accepts an optional title, published filename, content type, and display mode. |
| `screenshot_artifact` | Renders a local HTML, SVG, or image file so the agent can inspect its appearance.                                                                                  |
| `fetch_artifact`      | Downloads an existing artifact into the workspace for reading or revision. Access follows the permissions of the user who started the run.                         |

For a visual result, ask the agent to inspect a screenshot before publishing.
Choose `inline` for a browser preview or `attachment` for a download-only file.
The agent can set the content type, such as `text/html` or `application/pdf`, to
match the file.

### Update an agent artifact

Continue with the same agent and provide the existing artifact link:

> Update this report with the latest numbers. Keep the same artifact URL rather
> than creating a separate report: \[paste the artifact link].

The agent passes the existing `artifact_id` to `publish_artifact`. This publishes
a new version at the same URL. Updating this way requires the same originating
agent and the same user who created the artifact.

Each published file can be up to **10 MB**.

## Create from a connected assistant

With [Runlayer MCP](/runlayer-mcp) connected to your assistant, you can publish a
standalone artifact without creating a Runlayer Agent. Your account needs
**Use Agents** or **Manage Agents** access.

For example:

> Save this analysis as a Markdown artifact in Runlayer and give me its link.

The assistant uses `create_artifact` with the filename and content. Text files use
`content_format="text"`; binary files use `content_format="base64"`. For HTML or
Markdown, specify `content_type="text/html"` or `content_type="text/markdown"`;
text input otherwise defaults to plain text. Creation requires confirmation
through the tool's `confirm=true` parameter.

Standalone artifacts appear in **Created by me**. Use **Share** on the artifact to
grant colleagues access.

### Read and revise standalone artifacts

| Tool                   | Purpose                                                                                                                                   |
| ---------------------- | ----------------------------------------------------------------------------------------------------------------------------------------- |
| `list_artifacts`       | Finds artifacts visible to you.                                                                                                           |
| `get_artifact`         | Reads an artifact's metadata and link.                                                                                                    |
| `get_artifact_content` | Reads the stored content. Large responses can be truncated; check the returned `truncated` field before using the content for a revision. |
| `update_artifact`      | Replaces the content and metadata of a standalone artifact you own.                                                                       |
| `edit_artifact`        | Replaces or appends text in a standalone artifact you own without resending the whole file.                                               |

Updates and text edits require `confirm=true`, preserve the artifact link, and
create a new version. These MCP editing tools work on standalone artifacts;
revise an agent-created artifact through its originating agent.

See [version history](/artifacts/manage#version-history) to review or restore an
earlier result.