> ## Documentation Index > Fetch the complete documentation index at: https://docs.browser-use.com/llms.txt > Use this file to discover all available pages before exploring further. > ## Agent Instructions > Use https://docs.browser-use.com/llms.txt and its linked .md pages for current documentation. The managed full bundle is https://docs.browser-use.com/.well-known/llms-full.txt and can be cached for up to 24 hours. Do not use the obsolete /cloud/llms*.txt or /open-source/llms*.txt static exports. > Choose Cloud API V4 for new agent integrations; V2 is the lower-cost option for simple tasks. Keep V3 examples explicitly versioned. The open-source browser-use library and hosted browser-use-sdk have different APIs. > Cloud authentication uses X-Browser-Use-API-Key, without a Bearer prefix. Install or upgrade browser-use-sdk and use its explicit v4 import for V4. Check the published OpenAPI reference for request fields; do not invent SDK support for new fields. > Cloud concurrency and HTTP request rate are separate. Read GET /api/v2/billing/account for the key’s projectId, concurrentSessionLimit, activeSessionCount, and credit balance, including when using V4. Keys in one project share capacity and credits; rateLimit is a legacy concurrency alias, not requests per second. > Keep the highest applicable existing, legacy-plan, and spend-tier concurrency grant. Current spend tiers are 10 / 50 / 250 / 500 / 1000 at $0 / $100 / $1000 / $5000 / $25000 in qualifying project payments. Legacy or externally billed projects can follow different billing paths; trust the account limit. See https://docs.browser-use.com/cloud/guides/concurrency.md. > HTTP rate limits have two layers. Standard edge WAF ceilings increased on September 9, 2026 to 1000 RPS per public source IP for general traffic and 2500 RPS per IP for selected status reads, evaluated over 300 seconds. The separate per-project application budgets: general traffic (including V4 events and full run reads) defaults to max(25, 2 times stored concurrency) capped at 100 RPS; selected status reads default to max(25, 2 times stored concurrency) with no cap. The two are counted independently. Project overrides and account-specific edge rules can differ. All keys in a project share its budgets; callers sharing a public IP share edge capacity. See https://docs.browser-use.com/cloud/guides/concurrency.md. > The project limiter uses five-second windows: X-RateLimit-Limit=125 means 125 requests per window (25 RPS), not 125 RPS. Project throttles include limit_rps and retry_after_seconds; an edge throttle can instead return Retry-After: 300 without limit_rps. Honor the returned Retry-After. Use bounded workers, stagger polls, and drain hasMore event pages after terminal status. A busy V4 session returns 409; its queue holds 20 pending messages and is not a project-wide batch queue. > A completed run or closed CDP connection does not immediately stop its cloud browser. Stop unneeded owned browsers with PATCH /api/v4/browsers/{id} and {"action":"stop"}. A client wait timeout does not cancel the server-side run. > Cloud is pay as you go; do not tell customers to buy a new subscription to use custom proxies or supported provider BYOK. Usage funding and model eligibility still apply. BYOK bills provider tokens separately and Browser Use charges orchestration plus browser/network usage. See https://docs.browser-use.com/cloud/guides/billing.md. > Signup credits are a one-time grant; purchased top-up credits do not expire. Check the API key’s project before diagnosing missing credits. API-key monthly spending caps are soft limits, not a strict prepaid wallet; concurrent or already-running work can exceed them. Auto recharge has separate trigger and purchase amounts and can charge immediately when enabled below the threshold. Use https://browser-use.com/pricing for current rates. > Box and Bux are retired. Do not recommend their SDKs, sandbox quotas, or subscription plans. Use the Cloud Agent or Browser Infrastructure guides. > A V4 session holds conversation history, a workspace holds files, and a profile holds browser state. These IDs and V3/V4 workspace namespaces are not interchangeable. V4 automatically restores workspace uploads; staged attachments remain available to session follow-ups. Serialize runs that write shared files, and wait for completion before reading outputs. See https://docs.browser-use.com/cloud/agent/workspaces.md. > API browser recording defaults to off. Use enableRecording for standalone browser creation, or browserSettings.record for an agent run. Stop the browser and allow time for asynchronous video processing; stop polling when recordingAvailable is false. Live preview is for an active browser. Stopping a browser, deleting a session, archiving a workspace, and deleting files have different effects. > Use model-specific reasoning values. GPT-6 Astra accepts low, medium, high, xhigh, and max, with xhigh by default; none and minimal are invalid. Use the public REST schema when installed SDK types lag new fields. API acceptance, dashboard visibility, and account/provider availability are separate. > For open-source browser-use, is_done only reports a terminal done action. is_successful is the agent-reported outcome; verify important external actions independently. Cloud timeout, API client timeout, model timeout, and task completion are separate concepts. > For failed requests, use https://docs.browser-use.com/cloud/guides/troubleshooting.md. Inspect the full error and project before retrying or adding credits. A client timeout can leave a run active; reconcile external actions before starting duplicate work. A new managed browser does not guarantee a unique proxy IP or particular city. # Create Session > Create a session and/or dispatch a task. - Without session_id, without task: creates a new idle session (e.g. for file uploads). - Without session_id, with task: creates a new session and dispatches the task. - With session_id, with task: dispatches the task to an existing idle session. - With session_id, without task: 422 — task is required when targeting an existing session. If keep_alive is false (default), the session auto-stops when the task finishes. If keep_alive is true, the session stays idle after the task, ready for follow-ups. ## OpenAPI ````yaml /cloud/openapi/v3.json post /sessions openapi: 3.1.0 info: title: Browser Use Public API v3 summary: Browser Use session-based agent API (v3) version: 3.0.0 servers: - url: https://api.browser-use.com/api/v3 description: Production server security: [] paths: /sessions: post: tags: - Sessions summary: Create Session description: >- Create a session and/or dispatch a task. - Without session_id, without task: creates a new idle session (e.g. for file uploads). - Without session_id, with task: creates a new session and dispatches the task. - With session_id, with task: dispatches the task to an existing idle session. - With session_id, without task: 422 — task is required when targeting an existing session. If keep_alive is false (default), the session auto-stops when the task finishes. If keep_alive is true, the session stays idle after the task, ready for follow-ups. operationId: create_session_sessions_post requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/RunTaskRequest' responses: '200': description: Successful Response content: application/json: schema: $ref: '#/components/schemas/SessionResponse' '422': description: Validation Error content: application/json: schema: $ref: '#/components/schemas/HTTPValidationError' security: - APIKeyHeader: [] components: schemas: RunTaskRequest: properties: task: anyOf: - type: string - type: 'null' title: Task description: >- The natural-language instruction for the agent to execute (e.g. "Go to amazon.com and find the best-rated wireless mouse under $50"). Required when dispatching to an existing session. model: $ref: '#/components/schemas/BuModel' description: >- The model to use. "gemini-3-flash" is fast and cheap, "claude-sonnet-5" is balanced, "claude-opus-4.7" is most capable (default), and "claude-opus-4.8" is the newest Opus-tier model. GPT-5.6 models are Browser Use native models when either direct OpenAI or Amazon Bedrock routing is configured; otherwise they require use_own_key=true. GPT-5.5 becomes native when Bedrock routing is enabled; otherwise it requires use_own_key=true. Other additional provider models (e.g. "gemini-3.5-flash" and "gpt-5.2") require use_own_key=true. See BuModel for details. default: claude-opus-4.7 thinkingLevel: anyOf: - $ref: '#/components/schemas/ThinkingLevel' - type: 'null' description: >- Optional model reasoning depth. Omit this field to preserve the model provider default. Supported values depend on the selected model: most supported Claude models and GPT-5.1+ models support disabled/low/medium/high; Gemini Flash models support all four (disabled maps to Gemini's minimal level); Claude Fable 5, earlier GPT-5 models, Gemini 2.5 Pro, o3/o4, and Grok support low/medium/high; Gemini 3.1 Pro supports low/high; GLM supports disabled/high. Unsupported model/level combinations are rejected. sessionId: anyOf: - type: string format: uuid - type: 'null' title: Sessionid description: >- ID of an existing idle session to dispatch the task to. If omitted, a new session is created. keepAlive: type: boolean title: Keepalive description: >- If true, the session stays alive in idle state after the task completes instead of automatically stopping. This lets you dispatch follow-up tasks to the same session, preserving browser state and files. default: false maxCostUsd: anyOf: - type: number - type: string pattern: ^(?!^[-+.]*$)[+-]?0*\d*\.?\d*$ - type: 'null' title: Maxcostusd description: >- Maximum total cost in USD allowed for this session. The task will be stopped if this limit is reached. If omitted, a default limit applies (capped by your available balance). When dispatching a follow-up task to an existing session (`sessionId` is set), supplying this value overrides the session's budget for the upcoming dispatch; otherwise the budget is automatically refreshed to current spend + default. profileId: anyOf: - type: string format: uuid - type: 'null' title: Profileid description: >- ID of a browser profile to load into the session. Profiles persist cookies, local storage, and other browser state across sessions. Create profiles via the Profiles API. workspaceId: anyOf: - type: string format: uuid - type: 'null' title: Workspaceid description: >- ID of a workspace to attach to the session. Workspaces provide persistent file storage that carries across sessions. Create workspaces via the Workspaces API. proxyCountryCode: anyOf: - $ref: '#/components/schemas/ProxyCountryCode' - type: 'null' description: >- Country code for the browser proxy (e.g. "US", "DE", "JP"). Set to null to disable the proxy. The proxy routes browser traffic through the specified country, useful for accessing geo-restricted content. default: us browserScreenWidth: anyOf: - type: integer maximum: 6144 minimum: 320 - type: 'null' title: Browserscreenwidth description: >- Custom browser screen width in pixels. Must be set together with browserScreenHeight. When omitted, the browser keeps its own default resolution. browserScreenHeight: anyOf: - type: integer maximum: 3456 minimum: 320 - type: 'null' title: Browserscreenheight description: >- Custom browser screen height in pixels. Must be set together with browserScreenWidth. When omitted, the browser keeps its own default resolution. outputSchema: anyOf: - additionalProperties: true type: object - type: 'null' title: Outputschema description: >- A JSON Schema that the agent's final output must conform to. When set, the agent will return structured data matching this schema in the `output` field of the response. Example: {"type": "object", "properties": {"price": {"type": "number"}, "title": {"type": "string"}}}. enableScheduledTasks: type: boolean title: Enablescheduledtasks description: >- If true, the agent can create scheduled tasks that run on a recurring basis (e.g. "every Monday morning, check my inbox and summarize new emails"). Scheduled tasks are tied to your project and persist beyond the session. Note: all scheduled tasks are visible project-wide, so avoid enabling this in multi-user setups where task isolation is needed. default: false sensitiveData: anyOf: - additionalProperties: type: string type: object - type: 'null' title: Sensitivedata description: >- Key-value pairs of sensitive data (e.g. passwords, API keys) that the agent can use via secure placeholders. Keys are exposed to the LLM; values are never shown. The agent uses `key` placeholders in browser_type_text to securely enter values. enableRecording: type: boolean title: Enablerecording description: >- If true, records a video of the browser session. The recording URLs will be available in the `recordingUrls` field of the session response after the task completes. default: false skills: type: boolean title: Skills description: >- If true, the agent generates and persists reusable skills from completed tasks (saved per-domain in the DB and auto-injected into future runs). Set to false to skip skill generation — useful for privacy-sensitive tasks or to avoid the extra LLM cost. default: true agentmail: type: boolean title: Agentmail description: >- If true, provisions a temporary email inbox (via AgentMail) for the session. The email address is available in the `agentmailEmail` field of the session response. Useful for tasks that require email verification or sign-ups. default: true codeMode: type: boolean title: Codemode description: >- When true, the agent returns structured output with `text` (summary) and `code` (validated Python source) fields instead of free-form text. default: false cacheScript: anyOf: - type: boolean - type: 'null' title: Cachescript description: >- Controls deterministic script caching. `null` (default): auto-detected — enabled when the task contains `@{{value}}` brackets and a workspace is attached. `true`: force-enable script caching even without brackets (caches the exact task). `false`: force-disable, even if brackets are present. When active, the first call runs the full agent and saves a reusable script. Subsequent calls with the same task template execute the cached script with $0 LLM cost. Requires workspace_id when enabled. Example: "Get prices from @{{https://example.com}} for @{{electronics}}". useOwnKey: type: boolean title: Useownkey description: >- If true, uses your own LLM API key (configured in project settings) instead of Browser Use managed keys. You pay your provider directly for inference; Browser Use charges a reduced orchestration fee (0.2× of provider list prices). If no key is configured for the model's provider, the request is rejected. default: false autoHeal: type: boolean title: Autoheal description: >- When cache_script is active, controls whether a lightweight LLM validates the cached script output. If the output looks incorrect (empty, error, wrong structure), the system automatically re-triggers the full agent to generate a new version of the script. Set to false to disable validation and always return the raw script output. default: true type: object title: RunTaskRequest description: >- Create a new session, dispatch a task, or both. - **No `sessionId` + no `task`**: creates an idle session (useful for uploading files before running a task). - **No `sessionId` + `task`**: creates a new session and immediately runs the task. - **`sessionId` + `task`**: dispatches the task to an existing idle session. - **`sessionId` + no `task`**: returns 422 — a task is required when targeting an existing session. SessionResponse: properties: id: type: string format: uuid title: Id description: Unique session identifier. status: $ref: '#/components/schemas/BuAgentSessionStatus' description: >- Current session lifecycle status. Progresses through: `created` (sandbox starting) → `idle` (ready, waiting for task) → `running` (task executing) → `stopped` / `timed_out` / `error`. Poll this field to track progress. model: $ref: '#/components/schemas/BuModel' description: The model tier used for this session. thinkingLevel: anyOf: - $ref: '#/components/schemas/ThinkingLevel' - type: 'null' description: >- Configured model reasoning depth for this session, or null when provider defaults are used. title: anyOf: - type: string - type: 'null' title: Title description: >- Auto-generated short title summarizing the task. Available after the task starts running. output: anyOf: - {} - type: 'null' title: Output description: >- The agent's final output. If `codeMode` was true, this will be an object with `text` (summary), `code` (Python source), and optionally `output` (execution result). If `outputSchema` was provided, this will be structured data conforming to that schema. Otherwise it may be a free-form string or null. outputSchema: anyOf: - additionalProperties: true type: object - type: 'null' title: Outputschema description: The JSON Schema that was requested for structured output, if any. stepCount: type: integer title: Stepcount description: Number of steps the agent has executed so far. default: 0 lastStepSummary: anyOf: - type: string - type: 'null' title: Laststepsummary description: >- Human-readable summary of the most recent agent step (e.g. "Clicking the Submit button"). Useful for showing real-time progress. isTaskSuccessful: anyOf: - type: boolean - type: 'null' title: Istasksuccessful description: >- Whether the task completed successfully. `true` if the agent achieved the goal, `false` if it failed or gave up, `null` if the task is still running or no task was dispatched. liveUrl: anyOf: - type: string - type: 'null' title: Liveurl description: >- URL to view the live browser session. Available immediately on session creation — can be embedded in an iframe to show the browser in real time. recordingUrls: items: type: string type: array title: Recordingurls description: >- URLs to download session recordings. Only populated on `GET /api/v3/sessions/{id}`; always `[]` in list responses. default: [] profileId: anyOf: - type: string format: uuid - type: 'null' title: Profileid description: ID of the browser profile loaded in this session, if any. workspaceId: anyOf: - type: string format: uuid - type: 'null' title: Workspaceid description: ID of the workspace attached to this session, if any. proxyCountryCode: anyOf: - $ref: '#/components/schemas/ProxyCountryCode' - type: 'null' description: >- Country code of the proxy used for this session, or null if no proxy. browserScreenWidth: anyOf: - type: integer - type: 'null' title: Browserscreenwidth description: >- Custom browser screen width set for this session, or null for the default. browserScreenHeight: anyOf: - type: integer - type: 'null' title: Browserscreenheight description: >- Custom browser screen height set for this session, or null for the default. maxCostUsd: anyOf: - type: string pattern: ^(?!^[-+.]*$)[+-]?0*\d*\.?\d*$ - type: 'null' title: Maxcostusd description: Maximum cost limit in USD set for this session. totalInputTokens: type: integer title: Totalinputtokens description: Total LLM input tokens consumed by this session. default: 0 totalOutputTokens: type: integer title: Totaloutputtokens description: Total LLM output tokens consumed by this session. default: 0 proxyUsedMb: type: string pattern: ^(?!^[-+.]*$)[+-]?0*\d*\.?\d*$ title: Proxyusedmb description: Proxy bandwidth used in megabytes. default: '0' llmCostUsd: type: string pattern: ^(?!^[-+.]*$)[+-]?0*\d*\.?\d*$ title: Llmcostusd description: Cost of LLM usage in USD. default: '0' proxyCostUsd: type: string pattern: ^(?!^[-+.]*$)[+-]?0*\d*\.?\d*$ title: Proxycostusd description: Cost of proxy bandwidth in USD. default: '0' browserCostUsd: type: string pattern: ^(?!^[-+.]*$)[+-]?0*\d*\.?\d*$ title: Browsercostusd description: Cost of browser compute time in USD. default: '0' totalCostUsd: type: string pattern: ^(?!^[-+.]*$)[+-]?0*\d*\.?\d*$ title: Totalcostusd description: Total session cost in USD (LLM + proxy + browser). default: '0' screenshotUrl: anyOf: - type: string - type: 'null' title: Screenshoturl description: >- Presigned URL of the latest screenshot (expires after 5 minutes). Only populated on `GET /api/v3/sessions/{id}`; always `null` in list responses. agentmailEmail: anyOf: - type: string - type: 'null' title: Agentmailemail description: >- Temporary email address provisioned for this session (via AgentMail). Only present if `agentmail` was enabled. integrationsUsed: items: type: string type: array title: Integrationsused description: >- List of integration providers used during this session (e.g. ["gmail", "slack", "agentmail"]). createdAt: type: string format: date-time title: Createdat description: When the session was created. updatedAt: type: string format: date-time title: Updatedat description: When the session was last updated. type: object required: - id - status - model - createdAt - updatedAt title: SessionResponse description: >- Represents a session and its current state. Poll this endpoint to track task progress. The `status` field indicates the session lifecycle stage, and `output` contains the agent's structured result once the task completes. HTTPValidationError: properties: detail: items: $ref: '#/components/schemas/ValidationError' type: array title: Detail type: object title: HTTPValidationError BuModel: type: string enum: - bu-mini - bu-max - bu-ultra - gemini-3-flash - claude-opus-4.6 - claude-opus-4.7 - claude-sonnet-5 - claude-opus-4.8 - gpt-5.4-mini - glm-5.2 - minimax-m3 - grok-4.6 - glm-5.3-flash - deepseek-v4-flash-vision - claude-haiku-4.5 - gpt-5.2 - gpt-5-mini - gpt-5.5 - gpt-5.6-sol - gpt-5.6-terra - gpt-5.6-luna - gemini-3-pro - gemini-3.1-pro - gemini-3.5-flash title: BuModel description: >- The model to use for the agent. Each model has different capabilities and pricing. - `bu-mini` / `gemini-3-flash`: Gemini 3 Flash — fast and cost-effective. Best for simple, well-defined tasks like form filling or data extraction. - `bu-max` / `claude-sonnet-4.6` (legacy aliases) / `claude-sonnet-5`: Claude Sonnet 5 — balanced performance. Best for multi-step workflows that require reasoning and decision-making. - `bu-ultra` / `claude-opus-4.6`: Claude Opus 4.6 — capable general-purpose Opus tier. - `claude-opus-4.7`: Claude Opus 4.7 — most capable. Best for complex tasks that require advanced reasoning, long-horizon planning, or handling ambiguous instructions. - `claude-opus-4.8`: Claude Opus 4.8 — most capable Opus-tier model, state-of-the-art on long-horizon agentic work. - `gpt-5.4-mini`: GPT-5.4 mini — OpenAI's fast and efficient model. Best for tasks that benefit from OpenAI's capabilities. - `glm-5.2`: Z.ai GLM 5.2 — capable, low-cost open model. Best for cost-sensitive agentic tasks. - `minimax-m3`: MiniMax M3 — fast, very low-cost model. Best for simple, high-volume tasks. - `grok-4.6`: xAI Grok 4.6. Fast frontier model with a large context window. - `glm-5.3-flash`: Z.ai GLM 5.3 Flash. Multimodal, very low cost, served by Fireworks. - `deepseek-v4-flash-vision`: DeepSeek V4 Flash Vision. Experimental, very low cost. Additional provider models (e.g. `gemini-3.5-flash`, `gpt-5.2`, and `gpt-5-mini`) are selectable only with `use_own_key=true`, where your own provider key serves them. The GPT-5.6 family is native when either Browser Use's direct OpenAI key or Amazon Bedrock route is configured. GPT-5.5 becomes native only through the Bedrock rollout. ThinkingLevel: type: string enum: - disabled - low - medium - high title: ThinkingLevel description: Provider-neutral model reasoning depth. ProxyCountryCode: type: string enum: - ad - ae - af - ag - ai - al - am - an - ao - aq - ar - as - at - au - aw - az - ba - bb - bd - be - bf - bg - bh - bi - bj - bl - bm - bn - bo - bq - br - bs - bt - bv - bw - by - bz - ca - cc - cd - cf - cg - ch - ci - ck - cl - cm - co - cr - cs - cu - cv - cw - cx - cy - cz - de - dj - dk - dm - do - dz - ec - ee - eg - eh - er - es - et - fi - fj - fk - fm - fo - fr - ga - gd - ge - gf - gg - gh - gi - gl - gm - gn - gp - gq - gr - gs - gt - gu - gw - gy - hk - hm - hn - hr - ht - hu - id - ie - il - im - in - iq - ir - is - it - je - jm - jo - jp - ke - kg - kh - ki - km - kn - kp - kr - kw - ky - kz - la - lb - lc - li - lk - lr - ls - lt - lu - lv - ly - ma - mc - md - me - mf - mg - mh - mk - ml - mm - mn - mo - mp - mq - mr - ms - mt - mu - mv - mw - mx - my - mz - na - nc - ne - nf - ng - ni - nl - 'no' - np - nr - nu - nz - om - pa - pe - pf - pg - ph - pk - pl - pm - pn - pr - ps - pt - pw - py - qa - re - ro - rs - ru - rw - sa - sb - sc - sd - se - sg - sh - si - sj - sk - sl - sm - sn - so - sr - ss - st - sv - sx - sy - sz - tc - td - tf - tg - th - tj - tk - tl - tm - tn - to - tr - tt - tv - tw - tz - ua - ug - uk - us - uy - uz - va - vc - ve - vg - vi - vn - vu - wf - ws - xk - ye - yt - za - zm - zw title: ProxyCountryCode BuAgentSessionStatus: type: string enum: - created - idle - running - stopped - timed_out - error title: BuAgentSessionStatus description: >- Session lifecycle status. Progresses through: created → idle → running → idle → ... → stopped / timed_out / error. - `created`: Sandbox is starting up. The session is not yet ready to accept tasks. - `idle`: Sandbox is healthy and waiting for a task. You can dispatch a task or upload files. - `running`: A task is currently being executed by the agent. - `stopped`: Session was stopped — either explicitly via the stop endpoint, or automatically after task completion (when `keepAlive` is false). - `timed_out`: Session was cleaned up due to inactivity timeout. - `error`: Sandbox failed to start or encountered an unrecoverable error. ValidationError: properties: loc: items: anyOf: - type: string - type: integer type: array title: Location msg: type: string title: Message type: type: string title: Error Type type: object required: - loc - msg - type title: ValidationError securitySchemes: APIKeyHeader: type: apiKey in: header name: X-Browser-Use-API-Key ````