# IvreetMeet — MCP server > IvreetMeet is an Israeli service for transcribing and summarizing Hebrew meetings with AI. Its MCP server gives an AI assistant read access to your own meetings — search, transcripts with speakers and timestamps, summaries, decisions and action items — and can upload a recording or send a bot to a call. > IvreetMeet (עברית מיט) הוא שירות ישראלי לתמלול וסיכום פגישות בעברית. שרת ה-MCP שלו נותן לעוזר AI גישה לפגישות של החשבון שלך: חיפוש, תמלול לפי דוברים וזמנים, סיכומים, החלטות ומשימות, וגם העלאת הקלטה ושליחת בוט לפגישה. Product site: https://ivreetmeet.co.il (Hebrew) · product facts for agents: https://ivreetmeet.co.il/llms.txt This page: https://mcp.ivreetmeet.co.il/ · developer guide: https://mcp.ivreetmeet.co.il/docs · updated 2026-09-20 ## Endpoint - MCP, streamable HTTP: `https://mcp.ivreetmeet.co.il/mcp` - Authorization: OAuth 2.1 with PKCE. Discovery: `https://mcp.ivreetmeet.co.il/.well-known/oauth-protected-resource/mcp`. Clients that publish a client-ID metadata document (Claude, Claude Code, ChatGPT, VS Code) are accepted as they are; others may register dynamically. Sign-in opens an IvreetMeet approval page in Hebrew. - REST API for code: `https://mcp.ivreetmeet.co.il/v1` (personal API key, created in the account settings). OpenAPI: `/v1/openapi.json`. ## Scopes an AI app can be granted - `meetings:read` — קריאת פגישות, תמלולים וסיכומים - `meetings:write` — העלאת הקלטות וסיכום מחדש - `bots:write` — שליחת בוט לפגישות ## Tools - `list_meetings` — List the user's meetings, newest first - `search_meetings` — Search ALL of the user's meetings at once: summaries, decisions, action items (owner, due date, priority, status), participants, quotes, the communication analysis and its scores, and the full transcript with speakers and times - `get_meeting` — Meeting metadata: title, status, duration, processing state, speakers, links - `get_summary` — The full AI summary JSON (executive summary, action items, decisions, insights, participants, health) - `get_transcript` — Transcript with speakers and timestamps - `get_action_items` — Action items extracted by the summary (task, assignee, due date, priority, status) - `get_decisions` — Decisions recorded in the summary - `get_speakers` — Speakers of a meeting with their labels, assigned names and speech time - `get_health` — Communication-health analysis: aggregated score, six dimensions, objective metrics, highlights and improvements - `get_processing_status` — Processing state of an uploaded meeting (status, stage, progress, message, is_complete) - `upload_meeting` — Upload a recording from a public URL (mp3, m4a, wav, mp4 and other common formats, up to 500 MB) and start processing - `join_meeting` — Send the IvreetMeet bot to a Zoom / Google Meet / Microsoft Teams link - `resummarize` — Regenerate the summary of a completed meeting, optionally with new instructions (transcript, speakers and health are kept) - `search` — ChatGPT deep-research style search over ALL of the user's meetings: {results:[{id,title,url,text}]} - `fetch` — ChatGPT deep-research style fetch of one meeting: the summary and the transcript ## Connecting - Claude (claude.ai / Desktop): Settings → Connectors → Add custom connector → the endpoint above. It then also appears in Claude Code. - Claude Code: `claude mcp add --transport http ivreetmeet https://mcp.ivreetmeet.co.il/mcp` - Codex CLI: `codex mcp add ivreetmeet --url https://mcp.ivreetmeet.co.il/mcp` then `codex mcp login ivreetmeet` - Gemini CLI: `gemini mcp add --transport http ivreetmeet https://mcp.ivreetmeet.co.il/mcp` - Cursor / VS Code: one-click links on https://mcp.ivreetmeet.co.il/ - ChatGPT: Settings → Connectors → Developer mode → Create, with the endpoint above. ## Scope and privacy - Every answer is limited to the organization of the account that approved the connection; there is no cross-account access. - The approval is per app, shown on an IvreetMeet page that names the app and where the approval is sent, and can be disconnected at any time from the account settings. The account password never reaches the connected app. - Meetings of registered users are not used to train models. Per-speaker scores are withheld unless the organization turns them on. ## Good asks for an assistant - "מה הוחלט בפגישה של אתמול, ומי אחראי על מה?" - "כל המשימות הפתוחות מהשבוע האחרון" - "מה נאמר על התקציב? תן קישור לדקה בתמלול" - "העלה את ההקלטה הזו וסכם לפי הנקודות שביקשתי"