Roadmap

Blop is honest about its scope. The framework and blop CLI are shipped and open source; the hosted control plane (the part that clusters failures, heals them, and watches production) is being built in the open and is only partly live. This page is the line between the two.

Available now

• Local browser testing. blop test launches a real browser and runs agent-driven tests from .blop.ts specs.
• Intent-driven DSL. agentTest, describe, and defineAgentTest, with Playwright-backed browser tools behind a safe boundary.
• Multiple reporters. JSON results, a JSONL event log, and JUnit XML.
• Provider choice. OpenRouter, OpenAI, Groq, xAI, Mistral, Cerebras, and NVIDIA over the OpenAI chat-completions protocol; Anthropic and Google via OpenRouter.
• CI-native runs. GitHub Actions metadata is captured automatically and the exit code fails the build on any failing test.
• Platform upload. When BLOP_PLATFORM_URL and BLOP_API_KEY are set, the runner uploads the BlopRunResult after a run.

In early access

These appear on the marketing site and in the product vision, but are not generally available yet. Treat them as previews, not guarantees.

• Failure clustering: identical failures merged into one signal across runs.
• Self-healing pull requests: an agent that opens a scoped, verified fix PR when a test breaks.
• Synthetic monitoring: any test scheduled as a recurring uptime and p95 probe against production.
• Deeper integrations: Slack, Linear, and Jira wired into the control plane.

Coming next

• Remote and containerized browsers: connect over CDP and run a blop/browser Docker image, so CI can spin the browser up itself.
• A standalone tool server: expose the browser tools over HTTP for external agents.
• CI-based result ingestion: per-run secrets so a pipeline can post results without a user session.

For the full internal milestone breakdown, see docs/product/roadmap.md in the repository.