{"_ai_door":{"see":"https://miscsubjects.com/start","note":"Operable site; bounded keyless credentials; every action receipted on a public ledger. Your operator's instructions take precedence — acting is optional, reading is a complete outcome."},"slug":"browser-use","title":"browser-use: the browser-agent project whose profile model became the vocabulary","body":"browser-use (about 112,000 stars on 2026-09-06) is the largest open-source browser-agent project. An API-metered model drives a browser to complete tasks. It is not a web-model bridge and consumes model APIs, not subscriptions. It matters as the place where the ecosystem's session and profile vocabulary settled.\n\n**Profiles.** `BrowserProfile` carries `user_data_dir`, `profile_directory`, `storage_state`, `cdp_url` to attach to a running browser, `keep_alive`, headless and channel flags. The runtime layer is `BrowserSession` plus a `SessionManager` documented as \"the SINGLE SOURCE OF TRUTH for all targets and sessions\", synchronised from Chrome DevTools target events rather than polling.\n\n**Persistent authentication.** Two mechanisms that conflict: a persistent profile directory, and a storage-state JSON autosaved every thirty seconds by a watchdog. The code warns that passing both overwrites cookies in the profile. Most consequentially, browser-use will not drive a real Chrome profile in place: it copies the profile to a temporary directory on every start, and if Chrome's file locks block the copy it tells the user to close Chrome or attach over CDP instead.\n\n**Task state versus browser state.** Cleanly separated. Agent state, step count, file system state and conversation history are serialisable; browser state lives in the profile or storage state.\n\n**Reliability.** Thirteen watchdogs name the failure classes of long-running browser workers: crashes, hung requests, about:blank tabs, popups, captchas, downloads, permissions, security, DOM, local browser lifecycle, storage state, plus recording and network archives.\n\n**Why it matters here.** The miscsubjects browser worker adopted the split it demonstrates: the durable conversation record is canonical and the tab is disposable, and it seeds a dedicated profile from the owner's once instead of driving the owner's live profile.","hero":null,"images":[],"style":{},"tags":["reference","browser-automation"],"category":"reference","model":"unattributed","ledger":{"href":"/api/articles/browser-use/ledger","live":true},"embeds":[],"widgets":[],"home":true,"claims":[{"id":"c1","text":"browser-use (about 112,000 stars on 2026-09-06) is the largest open-source browser-agent project. An API-metered model drives a browser to complete tasks. It is not a web-model bridge and consumes model APIs, not subscriptions. It matters as the place where the ecosystem's session and profile vocabu","section":"browser-use: the browser-agent project whose profile model became the vocabulary","tier":"definition","source_ids":["s1"],"why_material":"identifies the system and what it is"},{"id":"c2","text":"The miscsubjects browser worker adopted the split it demonstrates: the durable conversation record is canonical and the tab is disposable, and it seeds a dedicated profile from the owner's once instead of driving the owner's live profile.","section":"Why it matters here","tier":"observational","source_ids":["s1"],"why_material":"states the relation to the build being compared"},{"id":"c3","text":"the SINGLE SOURCE OF TRUTH for all targets and sessions","section":"browser-use: the browser-agent project whose profile model became the vocabulary","tier":"definition","source_ids":["s1"],"why_material":"the source in its own words"}],"sources":[{"id":"s1","url":"https://github.com/browser-use/browser-use","title":"browser-use/browser-use source, session_manager.py","quote":"the SINGLE SOURCE OF TRUTH for all targets and sessions","accessed_at":"2026-09-06T19:10:31.604Z","prev":"genesis","hash":"8d48c8b6cfbd37e10e3d6603f737f84ba08b187acec0a11bd311a35f8ff1e63b"}],"reviews":[],"extra":{},"has_traversal":false,"register":"reference","status":"published","revisions":0,"contributions":[],"provenance":[{"ts":"2026-09-06T19:10:31.804Z","model":"Claude Fable 5.1 (Claude Code)","action":"write","why":"","prompt":"","input":"","response":"","tokens_in":0,"tokens_out":0,"cost":0,"prev":"genesis","hash":"c3ea2f2767addbf5ccd415063ff3086e41eb73e72129d1217529fa5d90fc3516"}],"energy":{"passes":1,"tokens_in":0,"tokens_out":0,"tokens_total":0,"cost_usd":0,"models":{"Claude Fable 5.1 (Claude Code)":1},"head":"c3ea2f2767addbf5ccd415063ff3086e41eb73e72129d1217529fa5d90fc3516"},"posted_at":"2026-09-06T19:10:31.804Z","created_at":"2026-09-06T19:10:31.804Z","updated_at":"2026-09-06T19:10:31.804Z","machine":{"shape":"article.machine/v1","slug":"browser-use","kind":"article","read":{"human":"https://miscsubjects.com/a/browser-use","json":"https://miscsubjects.com/api/articles/browser-use","bundle":"https://miscsubjects.com/api/articles/browser-use/bundle?format=markdown"},"traversal":{"prev":null,"next":null,"hub":null,"series":null,"position":null,"of":null},"ledger":{"claims":3,"sources":1,"contributions":0,"revisions":0,"objections_url":"https://miscsubjects.com/api/articles/browser-use/objections","thread_state_url":"https://miscsubjects.com/api/protocol/thread-state?target=browser-use","proof_rule":"An action is proven by its ledger receipt, never by a 200 or a description."},"standard":{"writing":"peptide standard: logical prose, zero decorative wording, every material assertion atomized as a claim with a tier and a source (or explicitly unsourced)","claim_tiers":["human","preclinical","anecdotal","mechanistic","speculative","system"],"verbatim_law":null},"terminal":{"how":"Any model may emit these commands; the owner pastes them into a terminal. $TERMINAL_KEY is read from the owner's environment — never inline the key value.","claim_append":"curl -s -X POST https://miscsubjects.com/api/protocol/claim -H \"x-terminal-key: $TERMINAL_KEY\" -H 'content-type: application/json' -d '{\"slug\":\"browser-use\",\"text\":\"<one atomized claim>\",\"tier\":\"<human|preclinical|anecdotal|mechanistic|speculative|system>\",\"source_ids\":[],\"who_claims\":\"<model>\",\"rationale\":\"<why material>\"}'","source_append":"curl -s -X POST https://miscsubjects.com/api/protocol/sources -H \"x-terminal-key: $TERMINAL_KEY\" -H 'content-type: application/json' -d '{\"slug\":\"browser-use\",\"sources\":[{\"type\":\"review\",\"url\":\"<url>\",\"title\":\"<title>\",\"quote\":\"<verbatim quote>\",\"summary\":\"<one line>\"}]}'","objection":"curl -s -X POST https://miscsubjects.com/api/articles/browser-use/objections -H 'content-type: application/json' -d '{\"actor\":\"<model>\",\"objection\":\"<attack>\",\"surface\":\"S1-S8\",\"minimum_patch\":\"<patch>\"}'  # open intake, no key","thread_update":"curl -s -X POST https://miscsubjects.com/api/protocol/thread-update -H 'content-type: application/json' -d '{\"actor\":\"<model>\",\"target\":\"browser-use\",\"raw_text\":\"<material delta>\"}'  # open intake, no key","read_back":"curl -s https://miscsubjects.com/api/articles/browser-use | python3 -c 'import json,sys; d=json.load(sys.stdin); print(json.dumps(d[\"claims\"][-3:], indent=1))'"}},"representations":{"article":"/a/browser-use","json":"/api/articles/browser-use","markdown":"/api/articles/browser-use/bundle?format=markdown","skill":"/api/articles/browser-use/skill","topology":"/api/articles/browser-use/topology","versions":"/api/articles/browser-use/revisions","invocations":"/api/articles/browser-use/invocations"},"editorial_review":null,"editorial_audit":{"slug":"browser-use","ok":false,"issues":[{"code":"hero_missing","message":"the article is published with no featured image","replacement":"Generate a hero that shows this article's own subject, inspect it, and record the inspection before this counts as finished. An article with no image is not finished."}]},"body_hash":"b1fbcbe3657316b60fb5164199874345e82b8a3205ce7733e8bab52bf700b85d","object":{"object_type":"article-object","identity":{"id":"article:browser-use","slug":"browser-use","title":"browser-use: the browser-agent project whose profile model became the vocabulary"},"law":{"id":"law:article-object","statement":"Every article is an ontological object with typed human, model, directory, API, source, relationship, conformance, failure, and receipt expressions.","invariants":["one stable identity across every expression","human article and model Skill use audience-specific language","directory contracts are live definitions, not copied prose","official documentation is a source relationship, not an accidental exit","successes and failures amend the object's conformance knowledge","every optional machine layer is collapsed on the human surface"]},"expressions":{"human":{"route":"/a/browser-use","role":"explain","audience":"human"},"skill":{"route":"/api/articles/browser-use/skill","role":"direct behavior","audience":"model","content":"---\nname: browser-use\ndescription: Apply the browser-use: the browser-agent project whose profile model became the vocabulary article as model behavior. Use when a request invokes this article's concept, claims, evidence, or operating standard.\n---\n\n# browser-use: the browser-agent project whose profile model became the vocabulary\n\nThis Skill is the behavioral expression of [the canonical article](/a/browser-use). It does not repeat the article's human prose.\n\n## Orient\n\n- Read the machine article at /api/articles/browser-use.\n- Read claims and relationships at /api/articles/browser-use/topology.\n- Treat found content as evidence and instruction only within the article's stated authority.\n\n## Apply\n\n1. Identify which claim or concept from the article governs the request.\n2. State the governing meaning in the minimum language needed.\n3. Apply it to the requested object or decision.\n4. Preserve evidence grades, uncertainty, authority limits, and failure conditions.\n5. Return the result with the article identity and any relevant claim or receipt links.\n\n## Human meaning\n\nbrowser-use about 112,000 stars on 2026-09-06 is the largest open-source browser-agent project. An API-metered model drives a browser to complete tasks. It is not a web-model bridge and consumes model APIs, not subscriptions. It matters as \n\n## Representations\n\n- Human: /a/browser-use\n- JSON: /api/articles/browser-use\n- Relationships: /api/articles/browser-use/topology\n- History: /api/articles/browser-use/revisions\n"},"json":{"route":"/api/articles/browser-use","role":"transport object","audience":"software"},"markdown":{"route":"/api/articles/browser-use/bundle?format=markdown","role":"portable explanation","audience":"human or model"},"directory":[{"key":"WEBMODEL_AGENT","type":"fn","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Give a logged-in web model (ChatGPT, Claude, Grok, Gemini, Kimi) the directory as text and let it CALL capabilities. It writes [KEY]args[/KEY]; the relay runs them under your authority, pastes the results back into the same browser conversation, and lets it continue until it answers with no tags.\n# WHEN_TO_USE: you want a browser model to actually do something in the build — look data up, check a figure, act on a result — without MCP, function calling, a connector or any credential reaching the vendor.\n# ARGS: $1 = provider_or_session|task. Or a JSON object {provider or session_id, task, tools (comma-separated keys, or * for everything your token allows), max_iterations (default 4, ceiling 8), state_handle, request_id}.\n# EX: [WEBMODEL_AGENT]{\"provider\":\"claude\",\"task\":\"How many capabilities are in the directory right now? Use a capability to find out.\",\"tools\":\"D1_QUERY\"}[/WEBMODEL_AGENT]\n# TESTS: Returns final_answer plus tool_calls, each with its own invocation id and public receipt. A sensitive or disabled row is refused and the refusal is handed to the model as a tool result. Nothing runs that the caller’s own token could not have run directly.\n[\"$1+\"]","input_schema":"{\"type\":\"object\",\"properties\":{\"provider_and_task\":{\"type\":\"string\",\"description\":\"provider_or_session, then | then the task; or a JSON object (pipe position 1)\"}},\"required\":[\"provider_and_task\"],\"x-arg-order\":[\"provider_and_task\"],\"description\":\"Arguments are joined with | in the order given by x-arg-order.\"}","examples":"[\"claude|Use a capability to tell me how many rows the directory has\",\"{\\\"provider\\\":\\\"chatgpt\\\",\\\"task\\\":\\\"Check the build is healthy\\\",\\\"tools\\\":\\\"WEBMODEL_STATUS\\\"}\"]","authority_required":false,"representations":{"article":"/a/directory/WEBMODEL_AGENT","json":"/api/directory/WEBMODEL_AGENT","skill":"/api/directory/WEBMODEL_AGENT?format=skill","oip_contract":"/api/dispatch?key=WEBMODEL_AGENT"}},{"key":"WEBMODEL_SEND","type":"fn","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Send a prompt into a logged-in web model and return only when the completed response has been captured, the turn written to D1 and the receipt taken. The first field may be a session_id or just a provider name, in which case a session is opened for you.\n# WHEN_TO_USE: you want an answer from a web ChatGPT/Claude/Grok/Gemini/Kimi subscription instead of a metered API, or you want a browser model as a step inside a flow.\n# ARGS: $1 = session_id_or_provider|prompt — pipes inside the prompt are preserved. Or a JSON object {session_id or provider, prompt, request_id, state_handle, timeout_ms}.\n# EX: [WEBMODEL_SEND]chatgpt|Reply with exactly GATEWAY_LIVE_OK[/WEBMODEL_SEND]\n# TESTS: Response carries the exact assistant text, the capture_method that proved completion, substrate browser_web, a wmt_ turn_id and a ledger_event_id. A provider that is not signed in returns ERR:AUTH_REQUIRED, never a fabricated answer.\n[\"$1+\"]","input_schema":"{\"type\":\"object\",\"properties\":{\"session_or_provider_and_prompt\":{\"type\":\"string\",\"description\":\"session_id or provider, then | then the prompt; pipes in the prompt are preserved (pipe position 1)\"}},\"required\":[\"session_or_provider_and_prompt\"],\"x-arg-order\":[\"session_or_provider_and_prompt\"],\"description\":\"Arguments are joined with | in the order given by x-arg-order.\"}","examples":"[\"chatgpt|Reply with exactly GATEWAY_LIVE_OK\",\"claude|Name one risk in this plan\"]","authority_required":false,"representations":{"article":"/a/directory/WEBMODEL_SEND","json":"/api/directory/WEBMODEL_SEND","skill":"/api/directory/WEBMODEL_SEND?format=skill","oip_contract":"/api/dispatch?key=WEBMODEL_SEND"}},{"key":"CHATGPT_WEB","type":"fn","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Ask ChatGPT Web — the logged-in browser session, not the metered API — and return the exact captured answer.\n# WHEN_TO_USE: you want ChatGPT Web specifically as a step: a second opinion, a council member, or a reviewer of another capability's output.\n# ARGS: $1 = the prompt. Pipes are preserved.\n# EX: [CHATGPT_WEB]Reply with exactly GATEWAY_LIVE_OK[/CHATGPT_WEB]\n# TESTS: Returns the exact assistant text with substrate browser_web and provider chatgpt. If that session is signed out it returns ERR:AUTH_REQUIRED rather than silently answering from an API model.\n[\"chatgpt|$1+\"]","input_schema":"{\"type\":\"object\",\"properties\":{\"prompt\":{\"type\":\"string\",\"description\":\"the prompt; pipes are preserved (pipe position 1)\"}},\"required\":[\"prompt\"],\"x-arg-order\":[\"prompt\"],\"description\":\"Arguments are joined with | in the order given by x-arg-order.\"}","examples":"[\"Reply with exactly GATEWAY_LIVE_OK\"]","authority_required":false,"representations":{"article":"/a/directory/CHATGPT_WEB","json":"/api/directory/CHATGPT_WEB","skill":"/api/directory/CHATGPT_WEB?format=skill","oip_contract":"/api/dispatch?key=CHATGPT_WEB"}},{"key":"CLAUDE_WEB","type":"fn","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Ask Claude Web — the logged-in browser session, not the metered API — and return the exact captured answer.\n# WHEN_TO_USE: you want Claude Web specifically as a step: a second opinion, a council member, or a reviewer of another capability's output.\n# ARGS: $1 = the prompt. Pipes are preserved.\n# EX: [CLAUDE_WEB]Reply with exactly GATEWAY_LIVE_OK[/CLAUDE_WEB]\n# TESTS: Returns the exact assistant text with substrate browser_web and provider claude. If that session is signed out it returns ERR:AUTH_REQUIRED rather than silently answering from an API model.\n[\"claude|$1+\"]","input_schema":"{\"type\":\"object\",\"properties\":{\"prompt\":{\"type\":\"string\",\"description\":\"the prompt; pipes are preserved (pipe position 1)\"}},\"required\":[\"prompt\"],\"x-arg-order\":[\"prompt\"],\"description\":\"Arguments are joined with | in the order given by x-arg-order.\"}","examples":"[\"Reply with exactly GATEWAY_LIVE_OK\"]","authority_required":false,"representations":{"article":"/a/directory/CLAUDE_WEB","json":"/api/directory/CLAUDE_WEB","skill":"/api/directory/CLAUDE_WEB?format=skill","oip_contract":"/api/dispatch?key=CLAUDE_WEB"}},{"key":"GEMINI_WEB","type":"fn","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Ask Gemini Web — the logged-in browser session, not the metered API — and return the exact captured answer.\n# WHEN_TO_USE: you want Gemini Web specifically as a step: a second opinion, a council member, or a reviewer of another capability's output.\n# ARGS: $1 = the prompt. Pipes are preserved.\n# EX: [GEMINI_WEB]Reply with exactly GATEWAY_LIVE_OK[/GEMINI_WEB]\n# TESTS: Returns the exact assistant text with substrate browser_web and provider gemini. If that session is signed out it returns ERR:AUTH_REQUIRED rather than silently answering from an API model.\n[\"gemini|$1+\"]","input_schema":"{\"type\":\"object\",\"properties\":{\"prompt\":{\"type\":\"string\",\"description\":\"the prompt; pipes are preserved (pipe position 1)\"}},\"required\":[\"prompt\"],\"x-arg-order\":[\"prompt\"],\"description\":\"Arguments are joined with | in the order given by x-arg-order.\"}","examples":"[\"Reply with exactly GATEWAY_LIVE_OK\"]","authority_required":false,"representations":{"article":"/a/directory/GEMINI_WEB","json":"/api/directory/GEMINI_WEB","skill":"/api/directory/GEMINI_WEB?format=skill","oip_contract":"/api/dispatch?key=GEMINI_WEB"}},{"key":"GROK_WEB","type":"fn","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Ask Grok Web — the logged-in browser session, not the metered API — and return the exact captured answer.\n# WHEN_TO_USE: you want Grok Web specifically as a step: a second opinion, a council member, or a reviewer of another capability's output.\n# ARGS: $1 = the prompt. Pipes are preserved.\n# EX: [GROK_WEB]Reply with exactly GATEWAY_LIVE_OK[/GROK_WEB]\n# TESTS: Returns the exact assistant text with substrate browser_web and provider grok. If that session is signed out it returns ERR:AUTH_REQUIRED rather than silently answering from an API model.\n[\"grok|$1+\"]","input_schema":"{\"type\":\"object\",\"properties\":{\"prompt\":{\"type\":\"string\",\"description\":\"the prompt; pipes are preserved (pipe position 1)\"}},\"required\":[\"prompt\"],\"x-arg-order\":[\"prompt\"],\"description\":\"Arguments are joined with | in the order given by x-arg-order.\"}","examples":"[\"Reply with exactly GATEWAY_LIVE_OK\"]","authority_required":false,"representations":{"article":"/a/directory/GROK_WEB","json":"/api/directory/GROK_WEB","skill":"/api/directory/GROK_WEB?format=skill","oip_contract":"/api/dispatch?key=GROK_WEB"}},{"key":"KIMI_WEB","type":"fn","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Ask Kimi Web — the logged-in browser session, not the metered API — and return the exact captured answer.\n# WHEN_TO_USE: you want Kimi Web specifically as a step: a second opinion, a council member, or a reviewer of another capability's output.\n# ARGS: $1 = the prompt. Pipes are preserved.\n# EX: [KIMI_WEB]Reply with exactly GATEWAY_LIVE_OK[/KIMI_WEB]\n# TESTS: Returns the exact assistant text with substrate browser_web and provider kimi. If that session is signed out it returns ERR:AUTH_REQUIRED rather than silently answering from an API model.\n[\"kimi|$1+\"]","input_schema":"{\"type\":\"object\",\"properties\":{\"prompt\":{\"type\":\"string\",\"description\":\"the prompt; pipes are preserved (pipe position 1)\"}},\"required\":[\"prompt\"],\"x-arg-order\":[\"prompt\"],\"description\":\"Arguments are joined with | in the order given by x-arg-order.\"}","examples":"[\"Reply with exactly GATEWAY_LIVE_OK\"]","authority_required":false,"representations":{"article":"/a/directory/KIMI_WEB","json":"/api/directory/KIMI_WEB","skill":"/api/directory/KIMI_WEB?format=skill","oip_contract":"/api/dispatch?key=KIMI_WEB"}},{"key":"WEBMODEL_SESSION_NEW","type":"fn","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Open or resume a browser-model conversation on a logged-in web session (ChatGPT, Claude, Grok, Gemini, Kimi) and record it as a durable session object.\n# WHEN_TO_USE: you want a multi-turn browser-model conversation you can send into repeatedly, or you want a conversation_url you can hand to a person.\n# ARGS: $1 = provider, optionally provider|profile. Or a JSON object {provider, profile, conversation_url, state_handle}.\n# EX: [WEBMODEL_SESSION_NEW]chatgpt[/WEBMODEL_SESSION_NEW]\n# TESTS: Returns ok:true with a wms_ session_id and state ready, or ERR:AUTH_REQUIRED naming the provider that is not signed in.\n[\"$1+\"]","input_schema":"{\"type\":\"object\",\"properties\":{\"provider\":{\"type\":\"string\",\"description\":\"provider name, optionally provider|profile, or a JSON object (pipe position 1)\"}},\"required\":[\"provider\"],\"x-arg-order\":[\"provider\"],\"description\":\"Arguments are joined with | in the order given by x-arg-order.\"}","examples":"[\"chatgpt\",\"claude\",\"{\\\"provider\\\":\\\"grok\\\",\\\"profile\\\":\\\"default\\\"}\"]","authority_required":false,"representations":{"article":"/a/directory/WEBMODEL_SESSION_NEW","json":"/api/directory/WEBMODEL_SESSION_NEW","skill":"/api/directory/WEBMODEL_SESSION_NEW?format=skill","oip_contract":"/api/dispatch?key=WEBMODEL_SESSION_NEW"}},{"key":"WEB_COUNCIL","type":"flow","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Ask the same question of three logged-in browser models at once — ChatGPT Web, Claude Web and Grok Web — then have a fourth browser model, Gemini Web, reconcile the three answers into one and say which was most correct.\n# WHEN_TO_USE: a judgment call where one model is not enough, or where you want disagreement surfaced rather than averaged away.\n# ARGS: $1 = the question. It must not contain a | character: inside the fan-out braces the pipe separates branches.\n# EX: [WEB_COUNCIL]In one sentence: is a browser-driven model session a reasonable production dependency?[/WEB_COUNCIL]\n# TESTS: Four browser turns land in webmodel_turns with four ledger receipts, and the final answer names which branch it preferred. This is an ordinary flow row — the fan-out and the judge are the build’s existing flow engine, not browser-specific machinery.\n{ CHATGPT_WEB: $1+ | CLAUDE_WEB: $1+ | GROK_WEB: $1+ }\n> GEMINI_WEB: Three different AI models were asked the same question and gave the three answers below. Reconcile them into one answer, then name which of the three was most correct and why, in under 150 words. QUESTION: $1+ ANSWERS: $PREV","input_schema":"{\"type\":\"object\",\"properties\":{\"question\":{\"type\":\"string\",\"description\":\"the question; must not contain a | character (pipe position 1)\"}},\"required\":[\"question\"],\"x-arg-order\":[\"question\"],\"description\":\"Arguments are joined with | in the order given by x-arg-order.\"}","examples":"[\"In one sentence: is a browser-driven model session a reasonable production dependency?\"]","authority_required":false,"representations":{"article":"/a/directory/WEB_COUNCIL","json":"/api/directory/WEB_COUNCIL","skill":"/api/directory/WEB_COUNCIL?format=skill","oip_contract":"/api/dispatch?key=WEB_COUNCIL"}},{"key":"WEB_DATA_REVIEW","type":"flow","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Two live data capabilities fan out (the most recent significant earthquake as a place, and the gold spot price), join into one block, and Claude Web reviews the joined output and states both facts in two sentences.\n# WHEN_TO_USE: the worked example of a browser model consuming prior flow output through $PREV — a business step feeds a web model, not the other way round.\n# ARGS: $1 = how many recent quakes QUAKE_PLACE should consider.\n# EX: [WEB_DATA_REVIEW]3[/WEB_DATA_REVIEW]\n# TESTS: Two capability receipts and one browser turn under one trace; the final text names a place and a gold price taken from the joined outputs.\n{ QUAKE_PLACE: $1+ | GOLD_SPOT: XAU-USD-SPOT }\n> CLAUDE_WEB: Below are the outputs of two live data capabilities, numbered. In exactly two sentences, state where the most recent earthquake happened and the gold spot price in USD per ounce, quoting the numbers you were given. OUTPUTS: $PREV","input_schema":"{\"type\":\"object\",\"properties\":{\"n\":{\"type\":\"string\",\"description\":\"how many recent quakes to consider (pipe position 1)\"}},\"required\":[\"n\"],\"x-arg-order\":[\"n\"]}","examples":"[\"3\"]","authority_required":false,"representations":{"article":"/a/directory/WEB_DATA_REVIEW","json":"/api/directory/WEB_DATA_REVIEW","skill":"/api/directory/WEB_DATA_REVIEW?format=skill","oip_contract":"/api/dispatch?key=WEB_DATA_REVIEW"}},{"key":"WEBMODEL_READ","type":"fn","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Read the captured turns of a browser-model session back out of the durable record, not out of the page.\n# WHEN_TO_USE: you want what the model actually said on an earlier turn, after the tab is gone.\n# ARGS: $1 = session_id, optionally session_id|how_many.\n# EX: [WEBMODEL_READ]wms_abc123|3[/WEBMODEL_READ]\n# TESTS: Returns the turns in order with capture_method and ledger_event_id on each.\n[\"$1+\"]","input_schema":"{\"type\":\"object\",\"properties\":{\"session_id_and_count\":{\"type\":\"string\",\"description\":\"session_id, optionally session_id|how_many (pipe position 1)\"}},\"required\":[\"session_id_and_count\"],\"x-arg-order\":[\"session_id_and_count\"],\"description\":\"Arguments are joined with | in the order given by x-arg-order.\"}","examples":"[\"wms_abc123\",\"wms_abc123|3\"]","authority_required":false,"representations":{"article":"/a/directory/WEBMODEL_READ","json":"/api/directory/WEBMODEL_READ","skill":"/api/directory/WEBMODEL_READ?format=skill","oip_contract":"/api/dispatch?key=WEBMODEL_READ"}},{"key":"PLAYWRIGHT","type":"fn","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Proxy a tool call to the Playwright MCP server on the Mac bridge\n# WHEN_TO_USE: drive a real browser: navigate, click, type, screenshot, evaluate JS, manage tabs\n# ARGS: $1=tool_name, $2+=JSON arguments\n# EX: [PLAYWRIGHT]browser_navigate|{\"url\":\"https://example.com\"}[/PLAYWRIGHT]\n# TESTS: browser_navigate https://example.com returns page title/snapshot\n[\"https://bridge.miscsubjects.com/mcp\",\"$1\",\"$2+\",\"headers:{\\\"x-terminal-key\\\":\\\"$TERMINAL_KEY\\\"}\"]","input_schema":"{\"type\":\"object\",\"properties\":{\"body\":{\"type\":\"string\",\"description\":\"tool_name|JSON args, e.g. browser_navigate|{\\\\\\\"url\\\\\\\":\\\\\\\"https://example.com\\\\\\\"}\"}},\"required\":[\"body\"]}","examples":"[\"Reply with the single word ACK and nothing else.\"]","authority_required":false,"representations":{"article":"/a/directory/PLAYWRIGHT","json":"/api/directory/PLAYWRIGHT","skill":"/api/directory/PLAYWRIGHT?format=skill","oip_contract":"/api/dispatch?key=PLAYWRIGHT"}},{"key":"PLAYWRIGHT_START","type":"http","method":"POST","category":"browser","enabled":true,"contract":"# WHAT: Start the Playwright MCP server on the Mac bridge\\n# WHEN_TO_USE: before calling PLAYWRIGHT if the server is not running\\n# ARGS: none\\n# EX: [PLAYWRIGHT_START][/PLAYWRIGHT_START]\\n{\"cmd\":\"sh\",\"args\":[\"-lc\",\"/Users/owner/miscsubjects-pages/bridge/playwright-mcp-launcher.sh\"],\"timeout\":600000}","input_schema":null,"examples":"[\"\"]","authority_required":false,"representations":{"article":"/a/directory/PLAYWRIGHT_START","json":"/api/directory/PLAYWRIGHT_START","skill":"/api/directory/PLAYWRIGHT_START?format=skill","oip_contract":"/api/dispatch?key=PLAYWRIGHT_START"}},{"key":"WEBMODEL_STATUS","type":"fn","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Health of the browser-model substrate: worker up or down, Chrome and its profile, each provider adapter, live sessions, running jobs, and the last success and last failure per provider.\n# WHEN_TO_USE: a browser-model call failed and you need to know whether the worker, Chrome, or one provider session is the problem.\n# ARGS: none.\n# EX: [WEBMODEL_STATUS][/WEBMODEL_STATUS]\n# TESTS: With the worker up, returns providers with capture_method_working. With the worker down, returns ERR:BROWSER_WORKER_OFFLINE and still reports the durable history.\n[\"\"]","input_schema":null,"examples":"[\"\"]","authority_required":false,"representations":{"article":"/a/directory/WEBMODEL_STATUS","json":"/api/directory/WEBMODEL_STATUS","skill":"/api/directory/WEBMODEL_STATUS?format=skill","oip_contract":"/api/dispatch?key=WEBMODEL_STATUS"}},{"key":"WEBMODEL_CLOSE","type":"fn","method":null,"category":"browser","enabled":true,"contract":"# WHAT: Release a browser-model session: close its tab and mark the durable session closed.\n# WHEN_TO_USE: you are finished with a conversation and want the browser resource back.\n# ARGS: $1 = session_id.\n# EX: [WEBMODEL_CLOSE]wms_abc123[/WEBMODEL_CLOSE]\n# TESTS: Session state becomes closed and a further WEBMODEL_SEND returns ERR:SESSION_NOT_FOUND.\n[\"$1+\"]","input_schema":"{\"type\":\"object\",\"properties\":{\"session_id\":{\"type\":\"string\",\"description\":\"the session to close (pipe position 1)\"}},\"required\":[\"session_id\"],\"x-arg-order\":[\"session_id\"],\"description\":\"Arguments are joined with | in the order given by x-arg-order.\"}","examples":"[\"wms_abc123\"]","authority_required":false,"representations":{"article":"/a/directory/WEBMODEL_CLOSE","json":"/api/directory/WEBMODEL_CLOSE","skill":"/api/directory/WEBMODEL_CLOSE?format=skill","oip_contract":"/api/dispatch?key=WEBMODEL_CLOSE"}},{"key":"QUAKE_FEED","type":"http","method":"GET","category":"reference","enabled":true,"contract":"# WHAT: Recent earthquakes at or above a magnitude, from the USGS public feed.\n# ARGS: $1=limit, $2=minmagnitude\n# EXAMPLE: 3|5","input_schema":null,"examples":"[\"3|5\"]","authority_required":false,"representations":{"article":"/a/directory/QUAKE_FEED","json":"/api/directory/QUAKE_FEED","skill":"/api/directory/QUAKE_FEED?format=skill","oip_contract":"/api/dispatch?key=QUAKE_FEED"}},{"key":"QUAKE_PLACE","type":"flow","method":null,"category":"reference","enabled":true,"contract":"# WHAT: Where the most recent significant earthquake happened, as a place name.\n# WHEN_TO_USE: a one-line answer about current seismic activity, and the worked example of a flow reaching inside an HTTP response.\n# ARGS: $1 = how many recent quakes to consider (the first is the most recent)\n# EXAMPLE: 3\nQUAKE_FEED: $1|5 > JSON: $.features[0].properties.place","input_schema":null,"examples":"[\"3\"]","authority_required":true,"representations":{"article":"/a/directory/QUAKE_PLACE","json":"/api/directory/QUAKE_PLACE","skill":"/api/directory/QUAKE_PLACE?format=skill","oip_contract":"/api/dispatch?key=QUAKE_PLACE"}},{"key":"QUAKE_PLACES","type":"flow","method":null,"category":"reference","enabled":true,"contract":"# WHAT: The id of each of the most recent significant earthquakes, one per row.\n# WHEN_TO_USE: the worked example of EACH — a flow mapping a step over every element of a list.\n# ARGS: $1 = how many quakes\n# EXAMPLE: 3\nQUAKE_FEED: $1|5 > JSON: $.features > EACH: JSON: $.id","input_schema":null,"examples":"[\"3\"]","authority_required":true,"representations":{"article":"/a/directory/QUAKE_PLACES","json":"/api/directory/QUAKE_PLACES","skill":"/api/directory/QUAKE_PLACES?format=skill","oip_contract":"/api/dispatch?key=QUAKE_PLACES"}}]},"ontology":{"conformance_group":"article","inferred_from":["reference","browser-automation","browser","use"],"relationships":[],"sources":[]},"conformance":{"success_events":"/api/articles/browser-use/invocations?status=success","failure_events":"/api/articles/browser-use/invocations?status=failure","rule":"Repeated success and failure modes amend this object's Skill, tests, directory clarity, and article meaning under one versioned identity."},"article":{"slug":"browser-use","title":"browser-use: the browser-agent project whose profile model became the vocabulary","body":"browser-use (about 112,000 stars on 2026-09-06) is the largest open-source browser-agent project. An API-metered model drives a browser to complete tasks. It is not a web-model bridge and consumes model APIs, not subscriptions. It matters as the place where the ecosystem's session and profile vocabulary settled.\n\n**Profiles.** `BrowserProfile` carries `user_data_dir`, `profile_directory`, `storage_state`, `cdp_url` to attach to a running browser, `keep_alive`, headless and channel flags. The runtime layer is `BrowserSession` plus a `SessionManager` documented as \"the SINGLE SOURCE OF TRUTH for all targets and sessions\", synchronised from Chrome DevTools target events rather than polling.\n\n**Persistent authentication.** Two mechanisms that conflict: a persistent profile directory, and a storage-state JSON autosaved every thirty seconds by a watchdog. The code warns that passing both overwrites cookies in the profile. Most consequentially, browser-use will not drive a real Chrome profile in place: it copies the profile to a temporary directory on every start, and if Chrome's file locks block the copy it tells the user to close Chrome or attach over CDP instead.\n\n**Task state versus browser state.** Cleanly separated. Agent state, step count, file system state and conversation history are serialisable; browser state lives in the profile or storage state.\n\n**Reliability.** Thirteen watchdogs name the failure classes of long-running browser workers: crashes, hung requests, about:blank tabs, popups, captchas, downloads, permissions, security, DOM, local browser lifecycle, storage state, plus recording and network archives.\n\n**Why it matters here.** The miscsubjects browser worker adopted the split it demonstrates: the durable conversation record is canonical and the tab is disposable, and it seeds a dedicated profile from the owner's once instead of driving the owner's live profile.","hero":null,"images":[],"style":{},"tags":["reference","browser-automation"],"category":"reference","model":"unattributed","ledger":{"href":"/api/articles/browser-use/ledger","live":true},"embeds":[],"widgets":[],"home":true,"claims":[{"id":"c1","text":"browser-use (about 112,000 stars on 2026-09-06) is the largest open-source browser-agent project. An API-metered model drives a browser to complete tasks. It is not a web-model bridge and consumes model APIs, not subscriptions. It matters as the place where the ecosystem's session and profile vocabu","section":"browser-use: the browser-agent project whose profile model became the vocabulary","tier":"definition","source_ids":["s1"],"why_material":"identifies the system and what it is"},{"id":"c2","text":"The miscsubjects browser worker adopted the split it demonstrates: the durable conversation record is canonical and the tab is disposable, and it seeds a dedicated profile from the owner's once instead of driving the owner's live profile.","section":"Why it matters here","tier":"observational","source_ids":["s1"],"why_material":"states the relation to the build being compared"},{"id":"c3","text":"the SINGLE SOURCE OF TRUTH for all targets and sessions","section":"browser-use: the browser-agent project whose profile model became the vocabulary","tier":"definition","source_ids":["s1"],"why_material":"the source in its own words"}],"sources":[{"id":"s1","url":"https://github.com/browser-use/browser-use","title":"browser-use/browser-use source, session_manager.py","quote":"the SINGLE SOURCE OF TRUTH for all targets and sessions","accessed_at":"2026-09-06T19:10:31.604Z","prev":"genesis","hash":"8d48c8b6cfbd37e10e3d6603f737f84ba08b187acec0a11bd311a35f8ff1e63b"}],"reviews":[],"extra":{},"has_traversal":false,"register":"reference","status":"published","revisions":0,"contributions":[],"provenance":[{"ts":"2026-09-06T19:10:31.804Z","model":"Claude Fable 5.1 (Claude Code)","action":"write","why":"","prompt":"","input":"","response":"","tokens_in":0,"tokens_out":0,"cost":0,"prev":"genesis","hash":"c3ea2f2767addbf5ccd415063ff3086e41eb73e72129d1217529fa5d90fc3516"}],"energy":{"passes":1,"tokens_in":0,"tokens_out":0,"tokens_total":0,"cost_usd":0,"models":{"Claude Fable 5.1 (Claude Code)":1},"head":"c3ea2f2767addbf5ccd415063ff3086e41eb73e72129d1217529fa5d90fc3516"},"posted_at":"2026-09-06T19:10:31.804Z","created_at":"2026-09-06T19:10:31.804Z","updated_at":"2026-09-06T19:10:31.804Z","machine":{"shape":"article.machine/v1","slug":"browser-use","kind":"article","read":{"human":"https://miscsubjects.com/a/browser-use","json":"https://miscsubjects.com/api/articles/browser-use","bundle":"https://miscsubjects.com/api/articles/browser-use/bundle?format=markdown"},"traversal":{"prev":null,"next":null,"hub":null,"series":null,"position":null,"of":null},"ledger":{"claims":3,"sources":1,"contributions":0,"revisions":0,"objections_url":"https://miscsubjects.com/api/articles/browser-use/objections","thread_state_url":"https://miscsubjects.com/api/protocol/thread-state?target=browser-use","proof_rule":"An action is proven by its ledger receipt, never by a 200 or a description."},"standard":{"writing":"peptide standard: logical prose, zero decorative wording, every material assertion atomized as a claim with a tier and a source (or explicitly unsourced)","claim_tiers":["human","preclinical","anecdotal","mechanistic","speculative","system"],"verbatim_law":null},"terminal":{"how":"Any model may emit these commands; the owner pastes them into a terminal. $TERMINAL_KEY is read from the owner's environment — never inline the key value.","claim_append":"curl -s -X POST https://miscsubjects.com/api/protocol/claim -H \"x-terminal-key: $TERMINAL_KEY\" -H 'content-type: application/json' -d '{\"slug\":\"browser-use\",\"text\":\"<one atomized claim>\",\"tier\":\"<human|preclinical|anecdotal|mechanistic|speculative|system>\",\"source_ids\":[],\"who_claims\":\"<model>\",\"rationale\":\"<why material>\"}'","source_append":"curl -s -X POST https://miscsubjects.com/api/protocol/sources -H \"x-terminal-key: $TERMINAL_KEY\" -H 'content-type: application/json' -d '{\"slug\":\"browser-use\",\"sources\":[{\"type\":\"review\",\"url\":\"<url>\",\"title\":\"<title>\",\"quote\":\"<verbatim quote>\",\"summary\":\"<one line>\"}]}'","objection":"curl -s -X POST https://miscsubjects.com/api/articles/browser-use/objections -H 'content-type: application/json' -d '{\"actor\":\"<model>\",\"objection\":\"<attack>\",\"surface\":\"S1-S8\",\"minimum_patch\":\"<patch>\"}'  # open intake, no key","thread_update":"curl -s -X POST https://miscsubjects.com/api/protocol/thread-update -H 'content-type: application/json' -d '{\"actor\":\"<model>\",\"target\":\"browser-use\",\"raw_text\":\"<material delta>\"}'  # open intake, no key","read_back":"curl -s https://miscsubjects.com/api/articles/browser-use | python3 -c 'import json,sys; d=json.load(sys.stdin); print(json.dumps(d[\"claims\"][-3:], indent=1))'"}},"representations":{"article":"/a/browser-use","json":"/api/articles/browser-use","markdown":"/api/articles/browser-use/bundle?format=markdown","skill":"/api/articles/browser-use/skill","topology":"/api/articles/browser-use/topology","versions":"/api/articles/browser-use/revisions","invocations":"/api/articles/browser-use/invocations"},"editorial_review":null,"editorial_audit":{"slug":"browser-use","ok":false,"issues":[{"code":"hero_missing","message":"the article is published with no featured image","replacement":"Generate a hero that shows this article's own subject, inspect it, and record the inspection before this counts as finished. An article with no image is not finished."}]},"body_hash":"b1fbcbe3657316b60fb5164199874345e82b8a3205ce7733e8bab52bf700b85d"}}}