Anthropic Alternative MCP Server

cancel_batch_message

Requests that have already been completed cannot be cancelled. Provide the batch ID. This is useful if you submitted a large batch by mistake and want to stop further processing to save costs. Cancel an in-progress batch message request

count_tokens

Requires the model ID and messages array. Returns the total input token count. Useful for estimating API costs and ensuring messages fit within context limits. Count tokens in a message before sending to Claude

create_batch_message

Each request in the batch has its own model, messages, max_tokens, etc. This is more cost-effective than individual requests when you have many independent prompts to process. Returns a batch ID for tracking. Use get_batch_message to check progress. Create a batch of message requests to Claude

get_batch_message

Returns the batch status (in_progress, succeeded, expired, canceling, canceled, failed), request counts (total, succeeded, errored) and individual results. Use the batch ID returned from create_batch_message. Get the status of a batch message request

list_models

Each model returns its ID (e.g. "claude-sonnet-4-20250514"), display name, creation date and capabilities. Use this to discover which models are available and their IDs for use with the send_message tool. List all available Anthropic Claude models

send_message

Requires the model ID (e.g. "claude-sonnet-4-20250514") and messages array in JSON format. Each message must have a "role" ("user" or "assistant") and "content" (text or array of content blocks). Optionally set max_tokens (default 1024), system prompt and temperature (0-1). Returns the assistant's response text. Send a message to Claude (Messages API)