Developer Integration
MCP Server
15 tools that let AI assistants record, transcribe, search, and create — all through the Model Context Protocol.
What is MCP?
The Model Context Protocol (MCP) is an open standard for connecting AI assistants to external data sources and tools. Instead of copying and pasting text into a chat window, MCP lets AI tools query your data directly through a structured interface.
SeaMeet implements an MCP server with 15 tools across 4 categories. AI clients can not only read your recordings — they can start recordings, take screenshots, search across all transcripts, and save generated content back to SeaMeet. Everything stays local.
15 Tools — Full Reference
Data & Read (5 tools)
seameet_list_recordings List recordings & screenshots with metadata. Filter by type, date range, and limit.
seameet_get_artifact Read a specific artifact: summary, transcription, action-items, chapters, key-decisions, description, or ocr-text.
seameet_get_asset_bundle Get the full manifest of all artifacts for a recording.
seameet_get_settings Get app settings: save directory, mic gain, recording format, language.
seameet_list_files List raw files in the recording save directory.
Write (2 tools)
seameet_save_artifact Save AI-generated content (summary, slides, etc.) alongside a recording.
seameet_rename_file Rename a recording or screenshot file.
Actions (6 tools) — require running app
seameet_start_recording Start audio or video recording. Supports source: "microphone", "screen", or "both".
seameet_stop_recording Stop the current recording. Returns the saved file path.
seameet_take_screenshot Capture a screenshot of the current screen.
seameet_recording_status Get current state: is recording active, paused, elapsed time, video included.
seameet_pause_recording Pause an active audio-only recording.
seameet_resume_recording Resume a paused audio recording.
Search (1 tool)
seameet_search_text Search across all artifact text (case-insensitive). Returns matching snippets.
Artifact Keys
Use these keys with {toolName} to retrieve specific data:
summarytranscriptionaction-itemschapterskey-decisionsdescriptionocr-text How it works
SeaMeet runs an MCP server locally
When the app is running, it spawns a local MCP server as a child process communicating over stdio. No network traffic leaves your device.
Configure your AI tool
Add SeaMeet to your MCP client config (Claude Code, Cursor, or any MCP-compatible tool). One line in your config file.
AI tools call SeaMeet directly
Your AI tool discovers all 15 tools automatically and can list recordings, read transcripts, start recordings, take screenshots, and save AI-generated artifacts — all through structured MCP tool calls.
How To Use
No manual configuration needed. SeaMeet handles everything automatically:
Run SeaMeet
When SeaMeet starts, it automatically spawns the MCP server and registers all 15 tools. No setup required.
Use the built-in AI Agent
Click the Agent tab in SeaMeet. The AI agent (GitHub Copilot or Claude) already has access to all 15 MCP tools — ask it to list recordings, search transcripts, or start a recording.
Or use Claude Code / Codex externally
SeaMeet auto-merges its MCP config into your Claude settings. Just run Claude Code in any terminal — it will discover SeaMeet's tools automatically while the app is running.
How it works under the hood: SeaMeet spawns the MCP server as a local child process (stdio transport). For tools that control the app (start recording, take screenshot), the MCP server calls back to the Electron main process via a local HTTP bridge, authenticated with a per-session secret that regenerates on each restart. Developers never need to configure ports or secrets manually.
Example usage
Once configured, your AI tool can interact with SeaMeet naturally:
You ask:
"List my recordings from this week"
AI calls:
seameet_list_recordings with dateFrom filter You ask:
"What were the action items from the product review?"
AI calls:
seameet_get_artifact with key: "action-items" You ask:
"Search all my meetings for mentions of the Q3 budget"
AI calls:
seameet_search_text with query: "Q3 budget" You ask:
"Start recording this meeting"
AI calls:
seameet_start_recording with source: "microphone" For developers: Full API reference including all 15 HTTP bridge endpoints, error codes, and authentication details is available at {url} when the app is running.
llms.txt — Machine-Readable Reference
SeaMeet publishes a {filename} file — a plain-text reference of every tool, endpoint, parameter, response shape, and error code. AI agents can read this file to understand SeaMeet's full API surface.