Coding tools (Claude Code, Cursor, VS Code)

Replace gt_live_XXXX with your API key. When you create or rotate a key on your Account page, one-click "Add to Cursor" and "Add to VS Code" buttons set everything up for you.

Rotating a key revokes the old one immediately and gives you a new key with the same name — use it if a key may have leaked.

Claude Code

claude mcp add --transport http gimmetext https://gimmetext.com/api/mcp --header "Authorization: Bearer gt_live_XXXX"

Cursor

~/.cursor/mcp.json (or .cursor/mcp.json in your project):

{
  "mcpServers": {
    "gimmetext": {
      "url": "https://gimmetext.com/api/mcp",
      "headers": { "Authorization": "Bearer gt_live_XXXX" }
    }
  }
}

VS Code

.vscode/mcp.json:

{
  "servers": {
    "gimmetext": {
      "type": "http",
      "url": "https://gimmetext.com/api/mcp",
      "headers": { "Authorization": "Bearer gt_live_XXXX" }
    }
  }
}

ChatGPT, Claude, Perplexity, Le Chat

No key needed: these connect by signing in. See Connect your AI.

Windsurf

mcp_config.json (open it from Cascade's MCP menu):

{
  "mcpServers": {
    "gimmetext": {
      "serverUrl": "https://gimmetext.com/api/mcp",
      "headers": { "Authorization": "Bearer gt_live_XXXX" }
    }
  }
}

Gemini CLI

~/.gemini/settings.json:

{
  "mcpServers": {
    "gimmetext": {
      "httpUrl": "https://gimmetext.com/api/mcp",
      "headers": { "Authorization": "Bearer gt_live_XXXX" }
    }
  }
}

Tools

ToolWhat it does
create_upload_linkReturns a single-use link (valid 30 minutes) where you drop a file while signed in.
list_jobsYour recent jobs, newest first (limit 1–100).
get_jobStatus, progress and error for one job.
get_transcriptTranscript as txt, srt or json. Long transcripts come in pages: pass offset and max_chars (1000–100000, default 50000); follow next_offset until it is absent.
get_usageYour plan and remaining minutes.

Prompts

PromptWhat it does
summarize_recordingSummarize my recording: a short summary plus key points.
meeting_action_itemsPull action items from a meeting: decisions, action items with owners if mentioned, and open questions.
lecture_study_notesTurn a lecture into study notes: an outline, key terms with definitions, and 5 review questions.

Each prompt accepts an optional job_id (UUID); omit it to use your latest finished recording. Discover them with prompts/list and request one with prompts/get. The response is a single user message guiding your AI to check get_job, then fetch every page with get_transcript before answering; it never embeds transcript text. Poll unfinished jobs about every 10 seconds until done or failed.

Example conversation

You: Give me an upload link.

AI: Here's your link: https://gimmetext.com/app/upload?t=XXXX (single use, 30 minutes).

Open it, drop your video or audio file, wait for it to finish.

You: Summarise my latest transcript.

AI: (calls list_jobs, then get_job, then every page of get_transcript) Here's a summary…

Transcript text is treated as information, never as instructions to the AI.

Also listed on

Questions? Start with the docs home or the machine-readable llms.txt.