Micro MCP server
Which servers in the official MCP registry are Micro's own, whether they answer, and what they offer. A program checked each of these, nobody copied them from a product page.
read 2026-10-19
- Registry namemu.micro/muverified in the MCP registry
The registry name sits under micro.mu. The registry only lets a publisher use that namespace after proving control of the domain through DNS or HTTP.
- Stateanswerschecked 2026-10-19
- Tools167
- Server calls itselfmicro
- Server version1.0.0
- Protocol version2025-03-26
On 2026-10-19 the server answered initialize and tools/list without a sign-in. First answer in 145 ms.
- Whether its tools do the product's job has not been judged yet.
- Checked on 2 of the 7 days it takes to judge reliability.
Once connected
- Keep approval on for chat_send, notify_send: they delete, send or pay without saying so.
- apps_create, apps_edit, blog_create, blog_delete, blog_update change data without a sign-in: anyone with the address can call them.
- Its definitions take about 17424 tokens of context: enable only the tools a task needs.
Said by rules from the measurements above. The likely first call is a model's pick from the tool list; read its description before relying on it.
- INITE MCP scoreafter a weeka score needs 7 daily checks: 2 so far
- Answeredmeasuring2 of the 7 daily checks it takes to print a share
- Median first answer561 ms
- Space in an agent's context≈ 17,424 tokensthe definitions of all its tools (167), which an agent carries on every turn
- Tool definitionsB · 89 of 100
- No description an agent can choose by: 1 of 167
- Parameters without a description: in 18 of 167
- Not saying whether it only reads or can destroy: 88 of 167
Worth a look
- Tools that delete, send or pay without telling the client so: chat_send, notify_send
- Tools that change data, on a server that answers without a sign-in: apps_create, apps_edit, blog_create, blog_delete, blog_update
A program checks the server once a day: initialize and tools/list, no model. Tokens are the definitions' length at about four characters a token. The grade averages six checks of every tool's name, description, schema and annotations; it says how well the tools are described, not whether they work. The score weighs what was measured and nothing else: answering (30%), doing the product's job (25%), definitions (20%), context (10%) and use (15%). A part that was not measured is left out, and a hidden instruction caps the score at 40. The state of vendors' MCP servers: method and numbers →
apps_buildQueue a durable app build and immediately return its ID and status URL. Use BuildStatus to retrieve progress and the completed app. Reuse request_key on retries. Builds survive server restarts
apps_buildsList your app build jobs and outcomes
apps_buildstatusRead your saved app build status by ID. Poll until complete or failed; the same ID survives restarts.
apps_collectionList your saved apps and their source references
apps_createCreate an app — a small, self-contained HTML tool hosted here. Takes the HTML; apps_build writes it for you from a description Apps run in a sandbox. Use fetch('/api/v1/<service>/<method>') or mu.service(name, method, args); mu.api does…
apps_editEdit an app you own — its name, description, tags, icon, HTML or price. Fields left out keep their value Apps run in a sandbox. Use fetch('/api/v1/<service>/<method>') or mu.service(name, method, args); mu.api does not exist. Direct…
apps_embedShow a saved micro app in the conversation. Embedding preserves access permissions and does not publish the app
apps_forkFork an app into your own account, to change independently of the original
apps_readRead an app or build status by slug, including source HTML for apps you own
apps_searchFind your saved private micro apps by name, description or tag
apps_testTest an app by checking its HTML and checking supported SDK calls; this is not browser execution, so an author finds out what is broken without opening it
archive_listWith a kind, the most recent entries of it. Without one, what kinds are archived and how much of each
archive_searchSearch everything this instance has collected, across news, video, markets and posts at once. Use it when the question crosses a service, or when you do not know which service would hold the answer; news_search is narrower and better when…
blog_createPublish a post to the caller's blog. For anything meant to be read later by other people — notes, write-ups, announcements. For a private note to yourself, prefer files or memory
blog_deleteDelete one of the caller's own blog posts, by id or title. Refuses posts written by anyone else, and refuses an ambiguous title rather than guessing. Irreversible, so confirm with the user first
blog_listRead recent blog posts — titles, snippets and ids
blog_readRead one blog post in full, by id or by title. Use after blog_list has found a candidate and the summary is not enough
blog_updateEdit one of the caller's own posts. Fields left out keep their current value
bookmarks_addSave an article, video, blog post or link privately. Re-saving preserves its note
bookmarks_annotateSet or clear a private note on one saved item
bookmarks_deleteRemove one of your saved items
bookmarks_getRead one saved item and its private note, with available archived text. Videos have descriptions, not transcripts
bookmarks_listFind your saved reading by text or kind, newest first. Private notes are searched too
browser_readOpen a page in a real browser and read it after its JavaScript has run. Use it when web_fetch comes back empty or with only a nav bar, which is what a page that builds itself in the browser looks like to a plain fetch. web_fetch is free…
browser_shotPhotograph a page and get back a URL for the picture. What it looks like, rather than what it says — a chart, a layout, a page whose content is an image. Ask for full to capture past the first screenful
chat_messagesRead the recent conversation in a discussion room
chat_roomsList discussion rooms that currently have activity
chat_sendSay something in a discussion room, as the caller. Use chat_rooms to find the room id
chat_sourceRead one stored message owned by the caller, by id
chat_sourcesList stored message ids owned by the caller
contacts_addSave someone to the address book. Adding a name already there updates it rather than making a second card
contacts_deleteRemove someone from the address book
contacts_findLook someone up in the address book by name, part of a name, or address. Use this before sending mail to a person named rather than addressed
contacts_listList everyone in the caller's address book
docs_deleteDelete one of your documents, by id
docs_listList your documents, most recently changed first, with an optional search over titles and bodies. Use this to find an id
docs_readRead one of your documents in full, by id or by exact title
docs_writeWrite a document — a title and a markdown body. Pass an id to replace one you already have. Private unless you set public. For something long enough to re-read; for a short thing to remember, use notes
events_createSchedule a reminder or event at a given time; optionally repeating, and optionally running a prompt through the agent when it fires
events_deleteCancel an event by id
events_freeFind when the caller has nothing booked — open slots of a given length, within working hours
events_listList the caller's upcoming events and reminders, each with its id
events_readRead one schedule owned by the caller
events_updateUpdate one of your scheduled events without changing its id. Omitted fields stay unchanged; empty notes, prompts or recurrence clear them
files_deleteDelete a file you own
files_drivereadRead the caller's connected Google Drive. Read-only; connect in Account first.
files_drivesearchSearch the caller's connected Google Drive. Read-only; connect in Account first.
files_getRead a stored file back by its id
files_listList the caller's stored files, newest first
files_putStore a file and get a URL for it — a report, a CSV, a transcript
files_shareMake a file readable by anyone with its URL, or private again
flights_airportReport what is happening at an airport right now: what is on the ground, what is on approach and what is climbing out. Live positions, not the timetable, so it says nothing about scheduled or delayed departures
flights_overheadList the aircraft flying near a location right now, nearest first, with altitude, speed, heading and distance. Live positions broadcast by the aircraft themselves, not a schedule
flights_statusScheduled, estimated and actual flight arrival/departure times from Aviationstack. Requires a configured provider.
flights_trackFind where an aircraft is right now by flight number ('BA117'), radio callsign ('BAW117') or registration ('G-ZBKL'). Only sees aeroplanes that are airborne and in range of a receiver — not finding one does not mean the flight was cancelled
food_hygieneFood hygiene ratings for UK businesses, from the Food Standards Agency inspections. Takes a business name, a town or postcode, or a lat/lon to look around
food_productLook up a packaged food by its barcode — name, brand, ingredients, allergens, nutrition per 100g, Nutri-Score and how processed it is
food_searchFind packaged foods by name or brand, with the barcode of each so one can then be looked up in full
forms_deleteDelete your form and its responses, or pass response_id to delete just one response.
forms_listList your forms, 50 at a time, with offset pagination.
The first 60 of 167, as the server listed them.
How to connect
Add the server to your MCP client config. Claude Desktop, Cursor and most other clients read this shape:
{
"mcpServers": {
"mu": {
"type": "http",
"url": "https://micro.mu/mcp"
}
}
}Add it
Claude Code, in a terminal
claude mcp add --transport http micro https://micro.mu/mcpclaude.ai, ChatGPT and other clients: add https://micro.mu/mcp as a custom connector in the settings.
Each line is something a program found when it went looking: the registry entry, the server's own answer, a spec that parses. A missing line means we did not find it, not that it does not exist.