LoginSign Up
← Back to all sparks
C

Currents

INFRA · APIS
Velocity7.5

Test results dashboard and analytics for Cypress and Playwright CI runs

currents.dev ↗

Currents starts recording any agent task, not just Playwright test runs.

playwrightagent-accessmcpagent-sessionsci-observability
◆Current state
Currents built a Playwright test dashboard and has spent the autumn opening it to agents: shareable failure links, scoped tokens, a hosted MCP server with OAuth, run attachments, a trace API and listings in the Claude and ChatGPT/Codex plugin directories. On November 12 it added Sessions, where an agent saves the browser recording or files from any task and gets one link with a page for people and a Markdown version for other agents.
◆Where it's heading
Until now every release made test results easier for agents to read. Sessions turns the direction around: agents write their own work into Currents, and the product's scope widens from CI test runs to a record of what agents did. The pattern of pairing a human page with a Markdown twin, first used for failure links in September, is now the product's standard output.
◆Prediction
Expect Sessions to be wired into the existing pull request comment and MCP server next, since earlier releases connected each new artifact to both within weeks.

◆Recent moves

  1. 2d ago

    Sessions: agents save task recordings to one shareable link

    ⚡ SPARK

    After months of making test failures readable to agents, Currents now lets agents store their own work: browser recordings or task files become a Session with a human page and a Markdown version. It is the first release aimed at agent output in general rather than Playwright runs.

    View source ↗
  2. 11d ago

    [New] Connect Currents from the Claude and ChatGPT/Codex plugin directories

    The hosted MCP server from September now appears in the Claude and ChatGPT/Codex plugin directories, with sign-in from the listing. It is distribution for the agent stack Currents built over two months, not a new capability.

    View source ↗
  3. 26d ago

    [Improved] A better GitHub run comment

    The pull request comment now updates during the run and shows each failed test with history and trace. It also hands developers and their agents a prompt to fix the failure, moving CI reporting toward repair.

    View source ↗
  4. 26d ago

    [New] Project Overview

    Projects now land on an Overview that reports health, trend direction and the top thing to fix. After six weeks of agent-facing access work, this is the human-facing triage layer on the same data; the note is brief and gives no detail on how priorities are ranked.

    View source ↗
  5. 1mo ago

    [New] Trace API for agents

    ⚡ SPARK

    The trace API is the deepest layer of the agent-access arc so far: after failure links, OAuth MCP and run attachments, agents can now navigate the full Playwright trace. It turns Currents' richest debugging artifact into something an agent can query.

    View source ↗
  6. 1mo ago

    [New] Run attachments

    Runs can now carry server logs, build output and merged reports, all collected on a new Attachments tab. That widens the failure context Currents holds, which is the raw material its MCP server and agent-readable failure links depend on.

    View source ↗