Skip to main content
Glama
brave-mcp
by menoxz

brave-mcp

Wrapper MCP local pour piloter le navigateur Brave (et Chrome/Edge) via chrome-devtools-mcp, avec gestion multi-navigateurs, sessions utilisateur réelles et reconnexion automatique.

Ce serveur ne crée pas de profil isolé : il se connecte au navigateur natif avec votre profil utilisateur (sessions, cookies, comptes déjà connectés).

Fonctionnement

start.js démarre un serveur MCP qui s'appuie sur la librairie officielle chrome-devtools-mcp (Puppeteer) pour les ~29 outils standards de navigation, et ajoute des outils de gestion du cycle de vie du navigateur :

Outil

Rôle

browser_list

Liste les navigateurs Chromium installés (Chrome, Brave, Edge) avec leur statut CDP

browser_launch

Lance un navigateur natif avec CDP et votre profil utilisateur

browser_use

Bascule la connexion vers un autre navigateur (sans redémarrage du serveur)

browser_status

Statut détaillé de la connexion CDP et pages ouvertes

browser_restart

Ferme un navigateur et le relance avec CDP + profil

Caractéristiques notables :

  • Sessions utilisateur : le navigateur est lancé avec --remote-debugging-port sur son profil réel — pas de profil jetable.

  • Port CDP dédié par navigateur : Chrome 9222, Brave 9223, Edge 9224.

  • Reconnexion à chaud : browser_use met à jour la cible sans redémarrer le serveur.

  • Health-check : toutes les 30 s, reconnexion automatique si le navigateur tombe.

  • Anti-corruption du TUI : toute sortie console.log/debug est redirigée vers stderr pour garder stdout propre pour le protocole MCP (JSON-RPC).

Related MCP server: chrome-remote-debugging-mcp

Prérequis

  • Node.js >= 20 (vérifié au démarrage par start.js)

  • Windows (détection des navigateurs via chemins standard, registre et where)

  • Chrome, Brave ou Edge installé (au moins un)

Installation

git clone https://github.com/menoxz/brave-mcp.git
cd brave-mcp
npm install
npm start

Configuration MCP

opencode

Ajoutez dans votre opencode.json (fichier de config utilisateur ou projet) :

{
  "mcp": {
    "web-browser": {
      "type": "local",
      "command": [
        "node",
        "C:\\chemin\\vers\\brave-mcp\\start.js",
        "--usageStatistics=false",
        "--performanceCrux=false"
      ],
      "environment": {
        "DEBUG": ""
      },
      "enabled": true
    }
  }
}

Claude Desktop / autres clients MCP

{
  "mcpServers": {
    "web-browser": {
      "command": "node",
      "args": ["C:\\chemin\\vers\\brave-mcp\\start.js", "--usageStatistics=false", "--performanceCrux=false"]
    }
  }
}

Lancement manuel du navigateur

Le serveur peut lancer le navigateur lui-même (browser_launch), mais vous pouvez aussi le démarrer à la main :

# Brave sur le port 9222 (profil utilisateur)
.\launch-brave.ps1

# Chrome, Brave ou Edge (détection automatique)
.\launch-browser.ps1 -Browser Brave

Outils exposés

  • Les outils standards de chrome-devtools-mcp (navigation, onglets, DOM, captures, etc. — ~29 outils).

  • Les outils browser_* listés ci-dessus.

Licence

MIT — Jean-Luc KOUMAGLO (menoxz), 2026.

Le serveur s'appuie sur chrome-devtools-mcp (Google), distribué sous licence Apache-2.0.

Available Tools

34 tools
browser_launchA

Lance un navigateur Chromium avec CDP et votre profil utilisateur (sessions, cookies, comptes). Si le navigateur tourne déjà sans CDP, il sera fermé et relancé avec CDP. Ne crée PAS de profil isolé — uniquement des sessions utilisateur.

Inputs: browser, confirmKill, additionalFlags.

browser_listA

Liste tous les navigateurs Chromium installés (Chrome, Brave, Edge) avec leur statut (en cours d'exécution, CDP actif).

Inputs: no named parameters.

browser_restartA

Ferme un navigateur en cours (sans CDP) et le relance avec CDP + votre profil utilisateur. Utile quand le navigateur est déjà ouvert mais pas en mode debug.

Inputs: confirmKill (required), browser.

browser_statusA

Affiche le statut détaillé de la connexion CDP et des pages ouvertes dans le navigateur.

Inputs: no named parameters.

browser_useA

Bascule la connexion vers un navigateur spécifique.

  • Noms: "Brave", "Chrome", "Edge" → navigateur NATIF (profil réel, sessions)

  • Ports: 9222, 9223 Si le navigateur natif n'a pas CDP, il sera fermé et relancé avec CDP + votre profil.

Inputs: target (required), confirmKill.

clickB

Clicks on the provided element

Inputs: uid (required), dblClick, includeSnapshot.

close_pageB

Closes the page by its index. The last open page cannot be closed.

Inputs: pageId (required).

dragB

Drag an element onto another element

Inputs: to_uid (required), from_uid (required), includeSnapshot.

emulateC

Emulates various features on the selected page.

Inputs: viewport, userAgent, colorScheme, geolocation, +2 more.

evaluate_scriptA

Evaluate a JavaScript function inside the currently selected page. Returns the response as JSON, so returned values have to be JSON-serializable.

Inputs: function (required), args, dialogAction.

fillB

Type text into an input, text area or select an option from a element.

Inputs: uid (required), value (required), includeSnapshot.

fill_formA

Fill out multiple form elements (inputs, selects, checkboxes, radios) at once. ALWAYS prefer this tool over multiple individual 'fill' or 'click' calls when interacting with forms. It is significantly faster, more reliable, and reduces turn count. Example: Fill username, password, and check "Remember Me" in one call.

Inputs: elements (required), includeSnapshot.

get_console_messageA
Read-only

Gets a console message by its ID. You can get all messages by calling list_console_messages.

Inputs: msgid (required).

get_network_requestA

Gets a network request by an optional reqid, if omitted returns the currently selected request in the DevTools Network panel.

Inputs: reqid, requestFilePath, responseFilePath.

handle_dialogA

If a browser dialog was opened, use this command to handle it

Inputs: action (required), promptText.

hoverB

Hover over the provided element

Inputs: uid (required), includeSnapshot.

lighthouse_auditA

Get Lighthouse score and reports for accessibility, SEO, best practices, and agentic browsing. This excludes performance. For performance audits, run performance_start_trace

Inputs: mode, device, outputDirPath.

list_console_messagesA
Read-only

List all console messages for the currently selected page since the last navigation.

Inputs: types, pageIdx, pageSize, includePreservedMessages.

list_network_requestsA
Read-only

List all requests for the currently selected page since the last navigation.

Inputs: pageIdx, pageSize, resourceTypes, includePreservedRequests.

list_pagesA
Read-only

Get a list of pages open in the browser.

Inputs: no named parameters.

new_pageB

Open a new tab and load a URL. Use project URL if not specified otherwise.

Inputs: url (required), timeout, background, isolatedContext.

performance_analyze_insightA
Read-only

Provides more detailed information on a specific Performance Insight of an insight set that was highlighted in the results of a trace recording.

Inputs: insightName (required), insightSetId (required).

performance_start_traceA

Start a performance trace on the selected webpage. Use to find frontend performance issues, Core Web Vitals (LCP, INP, CLS), and improve page load speed.

Inputs: reload, autoStop, filePath.

performance_stop_traceA

Stop the active performance trace recording on the selected webpage.

Inputs: filePath.

press_keyA

Press a key or key combination. Use this when other input methods like fill() cannot be used (e.g., keyboard shortcuts, navigation keys, or special key combinations).

Inputs: key (required), includeSnapshot.

resize_pageA

Resizes the selected page's window so that the page has specified dimension

Inputs: width (required), height (required).

select_pageA
Read-only

Select a page as a context for future tool calls.

Inputs: pageId (required), bringToFront.

take_memory_snapshotA

Capture a heap snapshot of the currently selected page. Use to analyze the memory distribution of JavaScript objects and debug memory leaks.

Inputs: filePath (required).

take_screenshotB

Take a screenshot of the page or element.

Inputs: uid, format, quality, filePath, +1 more.

take_snapshotA

Take a text snapshot of the currently selected page based on the a11y tree. The snapshot lists page elements along with a unique identifier (uid). Always use the latest snapshot. Prefer taking a snapshot over taking a screenshot. The snapshot indicates the element selected in the DevTools Elements panel (if any).

Inputs: verbose, filePath.

type_textA

Type text using keyboard into a previously focused input

Inputs: text (required), submitKey.

upload_fileB

Upload a file through a provided element.

Inputs: uid (required), filePath (required), includeSnapshot.

wait_forA
Read-only

Wait for the specified text to appear on the selected page.

Inputs: text (required), timeout.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3.1/5.0

    Scored across 34 tools

    Disambiguation3/5

    Several tools have overlapping purposes: fill, fill_form, and type_text all handle text input; browser_launch and browser_restart are nearly identical; browser_list and browser_status both report on browser state. While descriptions provide some clarification, an agent could easily misselect between these pairs.

    Naming Consistency3/5

    The majority of tools follow a verb_noun pattern (list_pages, close_page, take_screenshot, navigate_page), but several browser management tools reverse this (browser_list, browser_launch, browser_status) and some tools are single verbs (click, hover, fill, drag). The mixed ordering and occasional vague names (emulate) create an inconsistent but still readable pattern.

    Tool Count2/5

    With 34 tools spanning page interaction, form handling, performance tracing, network/console inspection, and browser management, the server is heavily overloaded. While the domain is broad, the count exceeds the 25+ threshold and many tools could be consolidated (e.g., merging fill and type_text, or combining browser_list/status).

    Completeness4/5

    The toolset covers a comprehensive browser automation lifecycle: navigation, page selection, interaction, form filling, file upload, screenshots, snapshots, performance traces, and network/console inspection. Minor gaps include lack of explicit cookie management or request interception, but these can be worked around via evaluate_script.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      C
      maintenance
      Unified MCP server for browser automation (Playwright + CDP + CLI) and bookmark management across Chrome, Firefox, Edge, and Brave, including cross-browser sync, deduplication, and a React dashboard.
      1
      MIT