BROWSER INFRASTRUCTURE · FOR AI AGENTS
Notte
Give agents hosted browser sessions with reusable authenticated state.
Use Notte’s operational MCP, APIs and SDKs to navigate pages, extract information and run browser workflows with managed session infrastructure.
Your integration checklist
- Capabilities
- What you need from your owner
- Create a Notte account and workspace, select a plan and authorize the MCP connection or issue an API key. Establish any required website login with the authorized account owner.
- Authentication
- The hosted MCP endpoint https://api.notte.cc/mcp supports OAuth workspace sign-in or Authorization: Bearer with a Notte API key. Keep saved login access limited to the intended workflow.
- Your first task
- Start one browser session, open a permitted page, extract a known field and stop the session after checking the result.
- Billing & payment
- Credits pay for browser sessions, functions, proxy bandwidth and model tokens. Plans set concurrency and session-duration limits; monthly plan credits expire each cycle, while purchased top-ups follow separate terms.
- Agent payment support
- Not verified. Programmatic usage does not establish autonomous checkout support.
- Operational limits
- Active sessions consume credits. Authentication may still be in progress when a session starts. Rate-window and maximum-active-resource errors both use 429 but have different error classes; capacity exhaustion can return 529.
ORIGINAL AI AGENT STORE INTEGRATION PLAYBOOK
Check authentication, task completion and browser shutdown separately
Editorial recommendation · Reviewed 2026-10-25
When to choose this service
Choose Notte when an agent needs operated browsers and authenticated website workflows through a direct tool interface.
Running browsers, preserving session resources and operating the hosting infrastructure require external capacity. Generating browser code does not provide those resources or the user’s login authority.
Suggested workflow
- Resolve the workspace and permitted website account. Start a controlled session and wait for authentication and active state before interacting.
- Perform one bounded task and validate the resulting page state. Record any submitted action so a reconnect cannot silently repeat it.
- Stop the session in success and failure paths. Check active-resource usage and credits before starting the next batch.
Acceptance checks before real work
- Verify the signed-in identity before acting on account data.
- Treat a browser action response separately from the website’s confirmed result.
- Confirm cleanup frees the session rather than waiting for a timeout.
If the workflow fails
Inspect the error class before retrying: close unused resources for active-session limits, back off for rate-window limits, and respect capacity retry guidance. Reconcile submitted actions before opening a replacement session.