Skip to content

Open-source AI browser agent for Chrome and Firefox (monorepo) 🧠

Merge main into d1-first-default-decision-model

Emre Sokulluab4f69f2614e
diff --cc CHANGELOG.md
index 41130beab,3fca454db..9dd02abed
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@@ -6,6 -6,14 +6,15 @@@ This changelog was generated from the r
## [Unreleased]
+ ### Changed
+ - The Apocalypse Mode button at the top of Settings now looks like the Providers and Support buttons beside it: a plain line icon instead of the yellow ☢, and no yellow outline while Apocalypse Mode is on. Hovering it still shows whether Apocalypse Mode is on.
++- The first-install guide's **Download local models** button now downloads D1 first and Ternary Bonsai 2 2B Vision after it, and the progress bar follows whichever is downloading. Bonsai follows even if you close the guide while D1 downloads, and if this GPU can't run D1, or D1's download stops, Bonsai starts anyway. Once D1 is downloaded, Auto uses it to check whether tasks and scheduled runs are finished even while WebBrain Cloud is your chat provider, so the guide no longer asks you to choose it in Settings. WebBrain Cloud's decision model still answers what a request asks for, and, as with any D1 answer, a "finished" from D1 is confirmed by the chat model. Chrome; the Auto change applies to Firefox too.
+
+ ### Fixed
+ - Settings search no longer looks as if its results belong to the tab you searched from. Searching "budget" on the CAPTCHA tab showed the Budgets settings while CAPTCHA was still highlighted. While a search shows its results, which can come from General, Budgets, CAPTCHA and Advanced, no tab is highlighted now, and the tab name above each group of results opens that tab.
+
+ ## [42.2.0] - 2026-10-20
+
### Added
- Mode picker in the composer's "!" menu: Auto stays the default, and you can choose Ask, Act or Dev there to keep that mode for the rest of the chat when Auto picks wrong. The "!" button then shows the mode (for example "âš  Act"), and a new chat starts in Auto again. The Hindi and Bengali names for Act, which read as "law", are corrected.
- Automatic mode selection: WebBrain now picks Ask, Act or Dev for each message, and the Ask, Act and Dev buttons are gone. It stays read-only unless your message asks it to do something in the browser. With a decision model configured (WebBrain Cloud includes one), the choice is made while the page is being read, so a message waits at most a quarter of a second longer. Without one, or when the decision model is unsure, the message starts in Ask; if it needs clicking, typing or opening pages, the model hands it over to Act, quoting your words. That hand-off counts only when your words, read again without the page, also ask for an action; the same message then restarts in Act with the usual plan review and permission prompts. Page content can never switch a message into Act, and right-click and selection prompts stay in Ask. Start a message with /ask, /act or /dev, or press Ctrl+Shift+A, X or D, to choose the mode for that message yourself. Text you selected on the page and the standalone window always stay in Ask.
diff --cc docs/security-model.md
index 3a29fa1e0,cb21bc1e1..3c655b5d0
--- a/docs/security-model.md
+++ b/docs/security-model.md
@@@ -220,8 -218,8 +220,8 @@@ The primary threat: a malicious page cr
| **Plan before Act** | When enabled, action-mode runs first produce a structured plan and wait for side-panel approval before any browser tool executes. In Try mode, planner JSON that remains invalid after repair degrades that turn to Ask/read-only; Strict stops. Scheduled runs can auto-approve the plan only through scheduler policy. |
| **Skill import boundary** | Skills can expose read-only HTTP tools and download-job tools through a `webbrain-tools` manifest. Importing or keeping the skill enabled is the trust decision for the declared HTTPS endpoint; declared skill tools use `credentials: "omit"` and should mark third-party results `resultPolicy: "untrusted"`. Download-job skill tools still require an action mode and the normal Downloads permission gate before saving files. |
| **WebMCP boundary** | Experimental WebMCP is on by default, so its tools and prompt guidance enter ordinary model requests unless the user opts out under Settings → Bridge. Chrome page-registered names, descriptions, schemas, frame URLs, annotations, outputs, and errors are page-controlled and always use the untrusted-content wrapper. Calls use opaque IDs. Ask may list tools but cannot invoke them. Because a callback can run arbitrary page logic, every invocation requires Act/Dev, fresh per-call confirmation, and a permission grant for the actual registration-frame origin; a page-authored `readOnly` hint never bypasses those gates. Missing/opaque frame identity fails closed, and the frame plus effective HTTP(S) security origin are revalidated immediately before dispatch to prevent navigation races from borrowing an old grant. |
- | **API mutation override** | A per-conversation `/allow-api` flag, or the default-on persistent setting under Settings → Advanced, *waives* the permission prompt for write-method network egress (`fetch_url`/`research_url` with POST/PUT/PATCH/DELETE). Neither option waives GET egress or any other capability. Conversation reset clears only the slash-command override. |
+ | **API mutation override** | A per-conversation `/allow-api` flag, or the default-on persistent setting under Settings → Advanced, *waives* the permission prompt for write-method network egress (`fetch_url`/`research_url` with POST/PUT/PATCH/DELETE) in Act and Dev. Ask stays read-only: an Ask run, including a turn that automatic mode starts in Ask, refuses a write-method request whatever these say. Neither option waives GET egress or any other capability. Conversation reset clears only the slash-command override. |
-| **Decision models** | Auto (default) uses the downloaded in-browser model for completion and scheduled checks, or Compass's managed route while Compass is the chat provider; Off uses none. In On, OpenRouter, direct TypeSafe and local/custom System One endpoints share independent per-feature switches. Keys are optional for local servers. Compass completion uses its managed decision route unless D1 or the in-browser chat model is chosen. Bounded/redacted task and fresh page evidence are untrusted data. Judges have no browser executor; ordinary tool calls still pass policy, permission and completion layers. Strict Secret Mode, cancellation, quota and cost limits apply. Unavailable or malformed judgments fall through to the active LLM and existing checks. |
+| **Decision models** | Auto (default) uses a downloaded D1 for completion and scheduled checks, whatever the chat provider; while Compass is the chat provider, Compass's managed route answers semantic decisions, and completion checks until D1 is downloaded. Off uses none. In On, OpenRouter, direct TypeSafe and local/custom System One endpoints share independent per-feature switches. Keys are optional for local servers. Compass completion uses its managed decision route unless D1 is downloaded (Auto) or chosen, or the in-browser chat model is chosen. Bounded/redacted task and fresh page evidence are untrusted data. Judges have no browser executor; ordinary tool calls still pass policy, permission and completion layers. Strict Secret Mode, cancellation, quota and cost limits apply. Unavailable or malformed judgments fall through to the active LLM and existing checks. |
| **`done()` verification** | Decision and tool-free LLM checks prefer vision, then AX. Confident pending/failed outcomes require read-only recovery before another submission. Accepted task-bound proof can satisfy generic evidence despite an unrelated comment form, while explicit recipient, payment, download, authorization and workflow contracts remain required. Existing dialog/form checks are the final fallback. |
| **Grouped search approval** | A same-site GET form with no credential, payment or hidden token field, and no side-effect action address, shares one task-scoped approval for its navigation, typing and submit when the page's structure marks it as a search: a search landmark, a search-typed field or one whose keyboard says search, code names such as `q` or `query`, a search address, or a submit control a model read, in any language, as only searching or filtering. Words in a field's visible label decide nothing. Every other form keeps its fresh confirmation. |
| **Duplicate-submit guard** | A second text click on the same caption within 45 seconds on the same tab and URL is blocked when a model read the caption, in any language, as committing something (sending, saving, paying, publishing, creating, confirming and so on). A caption read as anything else, or that could not be read, stays repeatable; app-observed popups, links, fields and draft tools are always repeatable. Both builds. |