SuperAPI AI Directory

seameet-mcp

Access SeaMeet meeting recordings, transcripts, AI summaries, screenshots, action items and webhooks, and control the desktop recorder.

Open website

About this tool

SeaMeet MCP server (seameet-mcp) is an MCP toolkit that lets AI agents record, transcribe, and mine meetings. Install once and use it in Claude Code, Claude Desktop, Codex, Antigravity, Cursor, OpenCode, GitHub Copilot CLI, Windsurf, and any MCP client.

[Purpose and Modes] The server automatically selects the richest available backend and exposes the union of both:

  • Desktop mode: When the SeaMeet desktop app is running, the agent gets the full recording toolset (start/stop recording, screenshots, live transcription, AI artifacts, search). No authentication required, and data never leaves the local machine.
  • Cloud mode: When the app is not running, the agent reads the synced library (recordings, transcriptions, summaries), can upload local audio/video files for transcription, and manages Webhooks over the network. Authorization is completed by logging into the web app, with no need to copy and paste keys, and it can be used in headless terminals.

tools/list returns a superset of the currently available tools; you can call seameet_status at any time to check the current mode.

[Main Capabilities]

  1. Recording and capture (desktop mode): seameet_start_recording (microphone, screen, or screen + system audio), seameet_stop_recording, seameet_pause_recording / seameet_resume_recording (audio-only pause/resume), seameet_recording_status, seameet_get_live_transcript (read live transcription during recording), seameet_take_screenshot.
  2. Process recording content (desktop = local files, cloud = synced library): seameet_list_recordings, seameet_list_files, seameet_get_settings; read generated artifacts (summary, transcription, SRT, chapters, action items, key decisions, screenshot OCR/description)—seameet_get_artifact; seameet_search_text searches artifact text; seameet_get_asset_bundle gets the asset bundle manifest; seameet_rename_file, seameet_save_artifact, seameet_regenerate_summary, seameet_list_templates.
  3. Cloud library (cloud mode): seameet_list_recent_recordings lists recent synced recordings (including files transcribed on app.seameet.ai); seameet_get_recording gets the summary, speaker-labeled transcription (paginated for long recordings), and chapters, as well as the action items and key decisions produced by the summary; the transcription is also available as the same SRT or VTT files downloaded from app.seameet.ai; seameet_get_recordings gets a certain portion (summary, action items, decisions, chapters) of up to 10 recordings at a time; seameet_transcribe_file uploads a local audio/video file (mp3, m4a, mp4, mov, wav, flac, ogg, oga, opus, webm, aac, amr, spx; up to 512 MB and 5 hours) and starts speaker-labeled transcription, immediately returns an assetId, and tells the agent when to poll seameet_get_recording (runs on local clients, so it works in Claude Code, Codex, Cursor, and Claude Desktop, but not in the web version claude.ai; uses your transcription quota); seameet_get_media_url gets a temporary media stream/download URL; seameet_search_recordings searches your recordings and recordings shared with you (full clips require Sync Pro); Meeting Memory—seameet_search_memory finds people, organizations, and topics, seameet_get_memory_page reads Memory pages with cited meetings; seameet_whats_changed summarizes 2–5 related recordings in chronological order (oldest first) with citations; seameet_rename_speakers renames speakers, seameet_regenerate_recording_summary regenerates the summary, seameet_rename_recording Rename recording; seameet_trash_recordings moves up to 20 recordings to the trash (recoverable within 30 days), seameet_restore_recordings restores them, seameet_list_recent_recordings (trashed: true) lists the trash (permanent deletion is still done on app.seameet.ai); seameet_get_usage checks storage usage, seameet_desktop_recording_help explains how to record with the desktop app; Webhook management—seameet_create_webhook, seameet_list_webhooks, seameet_update_webhook, seameet_delete_webhook (signed delivery on ai.ready / recording.synced; creation requires Sync Pro).
  4. Connection (any mode): seameet_status shows the connected mode, seameet_logout disconnects the cloud/switches accounts.

A total of 17 desktop tools + 19 cloud tools (fetched in real time, new versions appear automatically), plus 3 tools that run on the client (seameet_status, seameet_logout, seameet_transcribe_file). Tools that the current backend cannot provide return a clear structured error instead of failing silently.

[Hosted connector (no installation required)] SeaMeet also runs a hosted MCP server https://mcp.seameet.ai/mcp。将其作为连接器添加到 Claude or ChatGPT and sign in with a SeaMeet account (OAuth): no Node.js, no npx, no copying API keys. Any verified SeaMeet account can use it (including the free plan); Sync Pro unlocks full search snippets, the full Meeting Memory page, and Webhooks. Selection advice: chat apps and cloud-only use → hosted connector. Coding agents on the same machine as the SeaMeet desktop app, or uploading local files → this package.

[How to use] Every client runs the same command—npx -y @seameet/mcp—only the configuration format differs. See INSTALL.md for the complete per-tool recipes (mirrored at app.seameet.ai/mcp/install.md). Desktop mode prerequisites: install and run the SeaMeet desktop app—on macOS use brew install --cask seameet-ai/tap/seameet, on Windows use winget install seameet. Requires v3.2.0+; older running apps will report app_outdated. Cloud authorization (one-time, no copy-pasting keys): cloud mode is activated only when SEAMEET_API_KEY is provided or the device flow is completed. On the first cloud tool call with no key, the agent will display a short code + https://app.seameet.ai/link;你打开它(已登录)、点击 Authorize: a read-write key is minted and cached to ~/.seameet/credentials.json. After that it runs silently; it can be revoked at any time under the account's API keys. To disconnect or switch accounts, call seameet_logout (or rm ~/.seameet/credentials.json). Uploading large files: seameet_transcribe_file uploads in 8 MB chunks and sends stage notifications (notifications/progress: reading file, checking quota, each chunk start and completion, completing upload, starting transcription), at most once per second. If the host resets the tool timeout on progress it will keep waiting, but a stalled transfer sends nothing; hosts that do not reset may interrupt long uploads: in Claude Code you can raise the tool timeout for large files, e.g. MCP_TOOL_TIMEOUT=600000 claude (10 minutes). The transcription itself runs in the cloud after the tool returns, so only the upload counts. The file is opened once before sending, and each chunk is read from that open file. If it is rewritten, grows, or is replaced during upload, the upload stops with file_changed instead of sending a mix of old and new bytes. Recordings whose headers show more than 5 hours are rejected before sending.

[Claude Desktop one-click package and Claude Code plugin] manifest.json is the MCPB package manifest; the server is published as a one-click Claude Desktop extension: install via Settings → Extensions → Advanced → Install extension… and select dist/seameet.mcpb. Claude Desktop runs it with its bundled Node.js, no terminal required (desktop mode still requires the desktop app to be installed and running). This repository is also a Claude Code plugin marketplace. The plugin registers the MCP server and a /seameet skill that teaches the core workflows (recording+transcription, capturing bug reports, mining action items from meetings).

[Architecture and limitations] A thin stdio↔HTTP proxy. The SeaMeet desktop app hosts a local-only bridge (127.0.0.1:3741, fallback 3742/3743), authenticated with a random key per launch;on startup it writes credentials to $TMPDIR/seameet-mcp-bridge-<username>.json (permissions 0600), which this server reads and proxies (tools/list → GET /mcp-bridge/tools, tools/call → POST /mcp-bridge/call-tool). In desktop mode data does not leave the local machine. Cloud mode communicates over HTTPS with the hosted SeaMeet worker (the same server as the hosted connector), using the minted key. Every request sent to SeaMeet cloud carries User-Agent: seameet-mcp/<version>. Full tool reference (for LLMs): seameet.ai/llms.txt—also available locally when the app is running http://localhost:3741/llms.txt provides. Errors are machine-readable: tool failures return structured JSON for agents to branch on. Publishing automation—see PUBLISHING.md. License: MIT © SeaMeet.ai. Terms https://seameet.ai/en/terms/,隐私 https://seameet.ai/en/privacy/。