Yurei is a browser-automation tool announced on Show HN (2026-09-07) by creator memel06, pitched as 'like Claude in Chrome, but for any model and harness' — i.e. Anthropic-style in-browser agent control decoupled from Claude and from a specific agent harness. Two other independent builders have since shipped adjacent pitches: Brw (maxrev17, cross-harness automation with a remote-over-SSH workflow) and Browsentic (imshaikot, Apache-2.0, runs in the user's existing browser and bridges to agent CLIs like claude-code, codex, and cursor-cli instead of SDK APIs). The web search supplied with this stage returned only irrelevant motorcycle-forum results, so nothing on the open web corroborates any of these tools beyond the Show HN submissions themselves; all capability claims remain creator testimony with negligible traction (peak ~4.7 pts/h, no user success reports).
Scott's own wikis already carry the position this pitch applies: Model Perishability and Vendor Lock-In argue exactly for swap-ability over lock-in, and the browser-action territory is his own — the AI Chrome Vision implementation and BrowserUse-as-fallback in his scraping labs. Browsentic's CLI-bridge corroboration lifts the pattern's plausibility but adds no validated capability, no consequential party, and no adoption signal, so nothing here would change what he builds or argues; the one live thread is a cheap hands-on test of Browsentic against a harness like Ask, which stays an evaluation option rather than news.
ip:concept.model-perishabilityip:concept.vendor-lock-inip:source.breaking-new-ground-in-ai-agent-automation-ai-chrome-vision-and-openai-s-structured-output-ebookdev:technology.browser-useradar:claude-for-chrome-browser-agentradar:chrome-bridge-agent-browser-controlradar:concept.browser-agentsradar:concept.computer-use-agents
queries asked of Scott's wikis
- AI Chrome Vision browser action implementation notes
- Ask agent multi-provider harness browser tooling integration
- agent vendor lock-in model portability position
- model perishability replaceable-models claim
- computer use / browser automation evaluation attempts
- agent CLI bridge architecture claude-code codex cursor
now 0 pts/hpeak 13 pts/hcomments 0/hpeers p37momentum: steady2 platformsage 826h
points/hour across evidence · reading as of 2026-10-12 02:59:37.977291+11:00 · deterministic, not a model opinion
2026-10-11T02:33:59Z
Jshookmcp (CDP-based Chrome instrumentation via MCP) adds a fifth independent builder to the harness-agnostic browser-control pattern within ~a month, further corroborating builder-side convergence on the vendor-lock-in-removal direction. All five launches (Yurei, Brw, Browsentic, Rill, Jshookmcp) remain at negligible traction (scores 1–7, ~0 comments, case ~0.33 pts/h, 36th percentile, steady, 34 days old). The pattern is now clearly 'idea multiple builders reinvent weekly' with zero adoption or efficacy signal. Measured heat confirms cold episode despite hot neighbourhoods.
2026-10-11T02:33:19Z
evidence attached: hn.story.50039148 — CDP-based Chrome instrumentation MCP server extends the browser-agent tooling ecosystem across models/harnesses.
2026-10-06T10:23:03Z
Rill — a browser shell hosting Claude Code/Codex from any page — is a fourth independent implementation of harness-agnostic browser control within a month, so the pitch now reads as an idea multiple builders converge on unprompted; but a fourth zero-traction launch equally hardens the other half of the picture: broad builder interest with no validation or adoption anywhere. Direction corroborated more broadly, efficacy and lock-in reduction still unestablished.
2026-10-06T09:27:25Z
evidence attached: hn.story.49975836 — A second tool embedding cross-provider coding agents (Claude Code, Codex) into browsing, weak independent spread evidence for the browser-agent pattern.
2026-10-05T14:13:21Z
grounded: known/low — Scott's own wikis already carry the position this pitch applies: Model Perishability and Vendor Lock-In argue exactly for swap-ability over lock-in, and the bro
2026-10-05T14:03:50Z
Browsentic — an Apache-2.0 browser harness that bridges to existing agent CLIs (claude-code/codex/cursor-cli) instead of SDK APIs — is the second independent open-source implementation of the same pitch, lifting the case from a solo creator claim to a corroborated pattern: three independent builders (Yurei, Brw, Browsentic) now ship harness-agnostic, Claude-for-Chrome-style browser control. The corroboration is of the direction, not of any tool's efficacy — none has independent validation or user success reports, and traction on all three launches is negligible — so uncertainty stays high and heat low despite the hot browser-agents neighbourhood.
2026-10-05T13:28:23Z
evidence attached: hn.story.49964361 — Second independent open-source entrant pitching harness-agnostic, Claude-for-Chrome-style browser control, corroborating the seed case's vendor-lock-in-removal pattern.
2026-09-10T01:26:05Z
No new implementation evidence or user reports turn Yurei’s portability pitch into a validated workflow option. Brw remains a separate tool announcement, so the case is still a discovery lead rather than evidence of reduced browser-agent vendor lock-in.
2026-09-08T00:26:27Z
Brw is a separate creator-announced tool, not independent validation of Yurei; its attachment adds an adjacent portability pitch rather than corroborating this case. Brw’s claimed SSH workflow is a concrete evaluation lead, but Yurei’s compatibility and practical reduction of vendor lock-in remain unsupported.
2026-09-08T00:22:22Z
evidence attached: hn.story.49604261 — This released cross-harness browser automation tool materially corroborates the episode around model-agnostic browser agents, including remote execution.
2026-09-07T06:28:31Z
No substantive new evidence changes Yurei’s status as a tool-discovery lead. The echo repeats the creator’s announcement rather than independently validating portability, and no integration details or demonstrated workflow advantage are supplied.
2026-09-07T06:25:46Z
grounded: known/low — Yurei’s pitch repeats the replaceable-model position in Scott’s Model Perishability page and the browser-action territory of his AI Chrome Vision implementation
2026-09-07T06:23:07Z
case created — A creator-linked artifact makes browser-agent portability a bounded capability claim, though compatibility and reliability remain unsupported by the thin observation.