Skip to main content

Experimental

The Experimental section is where the PagerDuty FDE team publishes work that is actively being explored, tested, and refined — but is not yet part of the stable release.

Think of it as the lab. Things here may change without notice, break between updates, or eventually graduate into the main server. We share them openly so early adopters can try them, give feedback, and help shape what makes it into production.

Current Initiatives​

MCP Apps​

The first initiative under Experimental is MCP Apps — interactive HTML UIs that render directly inside agent chat sessions (VS Code, Claude Desktop, Goose, and other MCP-compatible clients).

MCP Apps work by embedding a sandboxed iframe in the chat interface. The app communicates bidirectionally with the MCP server: it can invoke tools to fetch or mutate data, and the server can push updates back to the UI in real time.

tip

MCP Apps is an ecosystem-wide extension to the Model Context Protocol, not a PagerDuty-specific feature. See the official MCP Apps documentation for the full spec.

AppDescription
Incident Command CenterAn interactive incident management dashboard — view, triage, and act on incidents without leaving your IDE

The experimental branch​

The primary source for experimental work is the experimental branch of the PagerDuty MCP Server repository:

github.com/PagerDuty/pagerduty-mcp-server/tree/experimental

This branch contains:

  • New tools under active development, not yet available in the published PyPI package
  • Prototype integrations exploring new PagerDuty API capabilities
  • Performance experiments such as alternative context strategies or streaming approaches
  • Configuration experiments that may inform future flags or options in stable releases

Running the experimental branch​

To run the server from the experimental branch directly:

uvx --from git+https://github.com/PagerDuty/pagerduty-mcp-server@experimental pagerduty-mcp

Or clone and run locally:

git clone -b experimental https://github.com/PagerDuty/pagerduty-mcp-server.git
cd pagerduty-mcp-server
uv run python -m pagerduty_mcp

For early adopters​

If you want to try experimental features before they ship:

  1. Run from the experimental branch using the commands above
  2. Expect occasional breaking changes between commits — this is not a stable surface
  3. Open issues or discussions on GitHub to share feedback directly with the team
  4. Watch the community streams where the FDE team often demos experimental work live before it's documented

Feedback​

The experimental branch exists because of community engagement. If something in the experimental branch is useful to you — or broken — the best place to say so is a GitHub issue tagged experimental.

caution

Experimental features are not covered by the standard support policy. Use them in non-production environments and treat them as subject to change at any time.