Autobrowserapproval MCP

What autobrowserapproval-mcp MCP server does

The autobrowserapproval-mcp MCP server exposes a hosted workflow for AI agent browser approvals. It is intended for AI agents and coding teams that need approval-related actions represented as structured JSON requests and responses. The service provides four MCP tools: request_page_approval, read_session_state, submit_approval_result, and export_approval_receipt.

These tools cover the documented approval lifecycle. An agent can request approval for a page, inspect the current session state, submit an approval outcome, and export a receipt for the completed approval process. The published description also identifies usage receipts and audit-ready workflow evidence as part of the hosted service.

How it works

The service is accessed remotely through the MCP endpoint at https://autobrowserapproval.clauxel.com/mcp. Clients communicate with it using Streamable HTTP and send JSON-RPC requests. Production requests must include a bearer token in the HTTP authorization header:

Authorization: Bearer <token>

The token is issued through the product website after selecting a plan and creating or requesting API access. A direct, unauthenticated browser visit to the MCP endpoint returns a JSON error rather than exposing internal service details.

The autobrowserapproval-mcp MCP server also publishes a server card at /.well-known/mcp/server-card.json. This gives compatible MCP clients a public discovery document for the hosted endpoint. The registry name listed by the project is com.clauxel.autobrowserapproval/autobrowserapproval-mcp.

Setup and configuration

To use the autobrowserapproval-mcp MCP server, open the product website, choose a plan, and obtain an API token. Add the remote MCP endpoint to a client that supports Streamable HTTP servers, then configure that client to send the token as a bearer credential. The README does not specify a required environment-variable name or a local installation package.

The repository is a documentation and directory-submission reference for the hosted product. It does not provide the private production source code, so deployment from the repository is not described or supported by the supplied material.

Tools and capabilities

The available MCP tools are:

• request_page_approval — starts or requests a page approval action.
• read_session_state — reads the state of an approval session.
• submit_approval_result — sends an approval result to the service.
• export_approval_receipt — exports a receipt for the approval workflow.

Together, these tools support requesting approval, checking progress, recording the decision, and obtaining workflow evidence. The service description also mentions usage receipts and audit-oriented records, but the supplied material does not define each tool's parameters, response schema, approval policy, or session-state fields.

Limitations and notes

The autobrowserapproval-mcp MCP server is a paid hosted service, not a self-contained local server. Production access depends on a product-issued bearer token and an active plan. The documentation does not list pricing details, token lifetime, rate limits, data-retention rules, supported browser engines, or detailed error behavior.

Client support depends on Streamable HTTP remote-server capability. The README does not name specific MCP client applications. The public endpoint, website, pricing page, server card, and registry entry are provided as hosted-service references rather than local runtime dependencies.