Skip to content

FAQ

A platform for measuring the agentic coding experience of your product across AI coding agents — by submitting requests that pair a task prompt, criteria, and a profile, and inspecting the structured pass/ fail reports of the resulting runs.

Teams who want to measure and compare AI coding agents across models, tools, and configurations on repeatable tasks.

Open the Portal URL provided for your deployment and follow its access requirements. See Access.

Three workers ship today:

  • GitHub Copilot CLIcoder-acp-copilot
  • Claude Code CLIcoder-acp-claude-code

See Coding agents & capabilities.

What’s the difference between a request and a run?

Section titled “What’s the difference between a request and a run?”

A request is what you submit (task prompt + criteria + profile). A run is one execution attempt of that request by a worker. The first attempt is created automatically; each retry adds another run on the same request. Logs and reports are per run.

So that “run profile X v3 again next month” gives you the same agent runtime — same agent version, same skills (pinned to commits), same extensions (pinned to versions), same MCP servers. Without that guarantee, year-over-year comparisons aren’t trustworthy.

If you need to change anything about a profile version, create a new version. The identity (name, description) is mutable separately.

Can I run the same task prompt against multiple profiles?

Section titled “Can I run the same task prompt against multiple profiles?”

Yes — that’s the main use case. Submit a request per profile against the same task prompt and criteria set.

On demand, not automatically. From the Portal, click Extract features on a task prompt. From the API, POST /api/v1/task-prompts/{id}/extract-features. See Working with prompt features.

Can I install VS Code extensions in any request?

Section titled “Can I install VS Code extensions in any request?”

ACP workers reject extensions with HTTP 400. See Choosing a coding agent.

Change its priority (range −10 to +10), or pause lower-priority requests. See Prioritizing & pausing requests.

Yes. The scope CLI lets you submit runs, stream logs, manage profiles, and more — all from the terminal. See Install the CLI to get started.

The auto-generated REST API reference is built from the committed OpenAPI snapshot. A narrative overview of the resource groups is at the REST API overview.