Rtkcontext MCP

What rtkcontext-mcp MCP server does

The rtkcontext-mcp MCP server connects an MCP client to a hosted OpenAI Codex context compressor. Its interface is organized around four operations: compressing context for a task, estimating potential context savings, issuing a context receipt, and reading compression history. The material describes the service as returning verdicts, receipts, usage logs, and audit-ready workflow evidence for AI agents and coding teams.

This is a remote service rather than a server implementation distributed in the repository. The repository serves as public documentation and a directory-submission reference; it does not include the private production source code.

How it works

The service is accessed through the Streamable HTTP MCP endpoint at https://rtkcontext.clauxel.com/mcp. An MCP client sends JSON-RPC requests to that endpoint, and production requests must include a bearer token in the Authorization header:

Authorization: Bearer <token>

The available operations are exposed as MCP tools and return structured JSON. The documented workflow is to obtain access, configure the remote endpoint in a compatible client, and then invoke the tools through MCP. A public server card is available at https://rtkcontext.clauxel.com/.well-known/mcp/server-card.json.

Setup and configuration

Using the rtkcontext-mcp MCP server requires a paid plan and a production bearer token. Start on the product website, choose a plan, and create or request an API token. Configure the MCP endpoint in a client that supports Streamable HTTP remote servers, then provide the token with each authenticated request according to that client’s configuration.

The documented product and pricing pages are hosted at https://rtkcontext.clauxel.com/. The registry name is com.clauxel.rtkcontext/rtkcontext-mcp. No local package installation or source build is described, because the repository does not contain the production service.

Tools and capabilities

The hosted endpoint exposes these tools:

• compress_context_for_task for the context-compression operation associated with a task.
• estimate_context_savings for estimating savings before or around compression.
• issue_context_receipt for producing a context receipt.
• read_compression_history for accessing compression history.

Together, these tools support compression workflows that need estimates, receipts, usage logging, or reviewable evidence. The supplied material does not document each tool’s parameters, response schema, compression algorithm, or supported context formats, so those details should be confirmed from the service’s current documentation or server card.

Limitations and notes

The rtkcontext-mcp MCP server is paid and requires a token for production calls. Unauthenticated browser requests to /mcp return a JSON error rather than internal service details. The README does not identify a local, offline, or self-hosted deployment option, and it does not name specific MCP clients. Compatibility therefore depends on whether the chosen client supports Streamable HTTP remote servers and bearer-token configuration.

The repository itself should not be treated as the implementation of the compressor. It documents the hosted endpoint, its public metadata, access model, and tool names. Pricing, token issuance, and service behavior may require checking the linked product site for current details.