Released version 1.0.0

The first release of AI Access MCP, an MCP server that lets your AI assistant (Claude Code, Cursor, Gemini CLI and others) hand part of the work over to another model: text to Gemini or GPT, images to OpenAI.

• generate_text – ask Gemini or GPT for a text; the server reads the files itself, so hundreds of kilobytes never pass through your assistant's context, and a long answer can go straight into a file
• Conversations – continue a call with its conversation and send just the new prompt; the model, instructions, schema and context carry over
• Gemini context cache – a batch of calls sharing a large input reads it at a tenth of the input price
• responseSchema – get structured JSON back, checked against your schema
• generate_image – draw a new image or edit up to 16 reference images into the file you name, with the prompt saved beside it and a preview for the assistant
• Cost – every result tells you what the call cost and how much the session has spent so far
• Safe by design – a request that may already have been paid for is never repeated, no file is ever overwritten, and files are read and written only under the roots you configure, with key files always off-limits