memobot-mcp
Unofficial MCP server for Memobot, enabling access to user profile, recordings, transcripts, summaries, and account usage through reverse-engineered API endpoints.
README
memobot-mcp
Personal, unofficial MCP server for Memobot, a voice-recording/ transcription app.
There is no public Memobot developer API. Everything here was reverse-engineered from
app.memobot.io's own network traffic against a real logged-in session. Endpoints, field
names, and the token scheme may change or break without notice.
Auth model
No password ever touches this codebase. The first time a tool needs a token and none is
cached, it pops a real Chromium window at Memobot's login page and waits for you to sign
in by hand (email/password, or the Google/Facebook/Apple buttons) — it just listens for
the app's own login response and reads the token out of it. That token is cached to
~/.config/memobot-mcp/credentials.json (0600 permissions) and reused for its ~1 year
validity. When it expires (or an API call gets a 401), the server first tries a silent
refresh via the cached refresh_token (POST /authen/api/v1/auth/token) — the browser only
pops up again if there's no cache at all or the refresh_token itself has been revoked
(e.g. after a password change).
Run standalone
uvx --from git+https://github.com/tinhvqbk/memobot-mcp memobot
Quick setup
Claude Code
claude mcp add memobot --scope user -- uvx --from git+https://github.com/tinhvqbk/memobot-mcp memobot
--scope user registers it in ~/.claude.json, so it's available in every project for
your user. Use --scope local to scope it to the current project instead.
Tools are namespaced as mcp__memobot__<tool>, e.g. mcp__memobot__get_current_user.
Every tool also has a matching MCP prompt, surfaced as a slash command
(/mcp__memobot__get_current_user, etc.) for a direct manual call.
Codex
codex mcp add memobot -- uvx --from git+https://github.com/tinhvqbk/memobot-mcp memobot
or in ~/.codex/config.toml:
[mcp_servers.memobot]
command = "uvx"
args = ["--from", "git+https://github.com/tinhvqbk/memobot-mcp", "memobot"]
OpenClaw
openclaw mcp add memobot --command uvx --arg --from --arg git+https://github.com/tinhvqbk/memobot-mcp --arg memobot
or in openclaw.json:
{
"mcp": {
"servers": {
"memobot": {
"command": "uvx",
"args": ["--from", "git+https://github.com/tinhvqbk/memobot-mcp", "memobot"]
}
}
}
}
Any other MCP client (Claude Desktop, etc.) uses the same command/args shape as
the Codex/OpenClaw JSON above.
Tools
get_current_user— account profilelist_recordings— paginated list of voice recordings/meetingsget_recording_transcript— raw detail (including transcript) for one recording, by audioId — see caveat belowget_recording_summary— raw AI-generated summary for one recording, by audioId — see caveat belowget_user_info— account usage/statsget_user_package— current subscription packageget_usage_stats— recording-minute usage for the current periodget_user_config— account-level feature flags/configget_api_key— the account's ASR/TTS API key and quotaget_notifications— recent in-app notifications
Known gaps
get_recording_transcript and get_recording_summary return raw JSON rather than a
parsed transcript/summary, because their response schema on a real recording is
unverified — a bespoke parser against an unconfirmed shape could silently produce wrong
output instead of an obvious error. The transcript reportedly lives under
content.document.children as paragraphs of word tokens (text + stime/etime in
ms); the summary lives in a content/text field; the audio detail also carries a
pre-signed, unauthenticated S3 URL under audio_document.url. Worth adding a parsed
variant once verified against real data.
File-upload and "join meeting" flows are unobserved and unimplemented.
Development
uv sync --all-groups # install runtime + dev dependencies
uv run ruff check . # lint
uv run ruff format . # format
uv run pytest -q # unit tests (mocked — no real network or browser calls)
CI runs lint, format-check, and tests on every push/PR to main. There's intentionally no
CI job that exercises the real Memobot API or the browser-login flow — that needs a real
account and a display, neither of which CI has.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.