@mgcrea/mcp-unifi-protect
A read-only-by-default MCP server for a UniFi Protect console, focused on cameras, recorded events, snapshots, footage export, and attached Protect devices.
Check authentication and reachability, force login with a 2FA code, or clear a cached session.
Inspect console system info, users, and saved live views.
List cameras, fetch a camera’s full configuration, and capture current snapshots.
List PTZ presets and patrols (no PTZ movement is available here).
Audit camera/system settings for misconfigurations and recording problems.
Search recorded events over time ranges with filters for event type, smart detections, camera, or configured location.
Get full event metadata and retrieve single or multiple event thumbnails for visual inspection.
Export recorded footage as MP4 files on disk.
List lights, sensors, viewers, and chimes with their current state and readings.
Use a raw GET escape hatch for undocumented private Protect API endpoints.
Mutating tools are absent from this schema; writes require
UNIFI_PROTECT_ALLOW_WRITES=1.
Provides tools for interacting with a self-hosted UniFi Protect console, including cameras, recorded events and smart detections, snapshots, video export, lights, sensors, viewers, chimes, and system information.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@@mgcrea/mcp-unifi-protect Show me motion events from the front door camera in the last hour"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
@mgcrea/mcp-unifi-protect
Model Context Protocol server for a self-hosted UniFi Protect console — cameras, recorded events and smart detections, snapshots, footage export, and the lights, sensors, viewers and chimes attached to it. Read-only by default: the tools that change anything are not registered at all unless you ask for them.
Features
Search recorded events over any time range — motion, person / vehicle / animal / package / licence-plate detections, doorbell rings — with each result already carrying its camera's name, not just an id, and what the camera saw: how many people, and each vehicle's plate, colour and type.
Snapshots and footage — capture a frame now, pull an event's thumbnail, export an MP4. All written to disk by default, so a still frame does not silently cost you a context window. An export says when its first frame was really recorded and whether it carries sound, so several cameras' footage can be lined up.
Devices — cameras, lights, sensors (with their temperature, humidity and light readings), viewers, chimes, live views and users.
Shaped responses. A console camera record is 8-15 KB of JSON; a list of ten is over 100 KB. List tools return the fifteen fields anyone actually asks about.
get_*returns everything.Stays up with no credentials, reporting what to configure through
unifi_protect_auth_statusrather than exiting and showing in your client as a bareConnection closed.
Related MCP server: UniFi MCP Server
Two ways to connect
|
| |
Reaches the console | directly on your LAN | via |
Credentials | host + username + password | API key + console id |
Auth mechanism | UniFi OS login → session cookie + CSRF token |
|
TLS | console's self-signed cert — pinned on first use | a real certificate, nothing to do |
Works off-LAN | no | yes |
Session state on disk | yes, mode | none |
Both modes expose exactly the same tools, because both speak the same private
Protect API — the connector forwards the whole /proxy/protect/... tree, the private
API included. That is not obvious and is worth stating plainly: Ubiquiti's official
Integration API has no historical query capability whatsoever, so if the connector only
carried that, cloud mode could not answer a single question about the past. It carries
the private API too, verified against a live console — bootstrap, events, cameras
and binary snapshots all answer 200.
So cloud mode is a full alternative, not a reduced one, and it removes the local account, the password, the session file, the CSRF handshake and the self-signed certificate problem in one go.
# cloud — no local account at all
UNIFI_PROTECT_API_KEY=… # unifi.ui.com → Settings → API Keys
UNIFI_PROTECT_CONSOLE_ID=… # curl -H "X-API-KEY: $KEY" https://api.ui.com/v1/hosts
# local — on the LAN
UNIFI_PROTECT_HOST=192.168.1.1
UNIFI_PROTECT_USERNAME=mcp
UNIFI_PROTECT_PASSWORD=…
# TLS needs nothing: the console's certificate is pinned on the first request.UNIFI_PROTECT_MODE is inferred as cloud when an API key and a console id are both
set, so it usually needs no setting. console/unifios/lan and
remote/site-manager/connector are accepted as synonyms, and an unrecognised value
is reported through unifi_protect_auth_status rather than killing the server.
Two traps in cloud mode. A key that works for /v1/hosts can still return
403 user cannot access host in the organization for a console outside the
organization it was issued in — valid key, wrong org, and the message reads nothing like
a credentials problem. And API keys are per-console: a key created on your Network
gateway or a UNAS does not authenticate against the NVR running Protect, and the NVR
rejects it exactly as it rejects a made-up key.
Why local mode needs a username and password
An API key looks like it ought to work here, and it is the obvious thing to reach for. It does not, and the reason is worth writing down so nobody spends an afternoon on it.
A key created on the console itself (UniFi OS → Control Plane → Integrations) is recognised — but only by Ubiquiti's official Integration API. It is refused by the private API this server depends on. Tested against a UNVR on Protect 7.2.105 with a key issued on that console:
Endpoint | With a console API key |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
A fabricated key returns 401 on the official API too, so the 200s above confirm the
key really was valid — the private API simply does not accept key auth.
Putting all three paths together:
Path | Authenticates with | Private API (event history, snapshots) |
| session cookie + CSRF | ✅ |
|
| ❌ |
|
| ✅ |
The asymmetry is not arbitrary. Over the connector, api.ui.com authenticates you by
key and then reaches the console over its own trusted channel, so the console is never
asked to accept a key on a private path. On the LAN there is no such intermediary, and
the private API only knows the session the web app itself uses.
So a local-only deployment needs a username and password. That is a property of Protect, not a shortcut taken here. Use a dedicated Local-Access-Only account with View Only rights, as described below, and the credential's blast radius stays small.
The one thing a console API key would unlock is the PTZ move commands
(ptz/goto, ptz/patrol/start, ptz/patrol/stop), which exist only on the official
Integration API — see Not implemented.
Security
Supply chain. Three runtime dependencies: the MCP SDK, zod, and
@mgcrea/unifi-protect — the console client,
which is ours and whose own dependencies are undici, ws and zod. There is no HTTP client
wrapper, no logger, no crypto library. That client used to be a copy inside this repo; the copy
and the original had already begun to drift, which is a bad way to hold knowledge about an
undocumented API. Published from CI with provenance via OIDC trusted publishing; the container
image is multi-arch, carries an SBOM, and is signed with cosign.
Your credentials. The username and password come from the environment or a config file, and
never leave this process except in the login request to your console. The resulting session
cookie is cached at ~/.config/unifi-protect/session.json with mode 600.
Certificate verification is ON by default, and now needs no setup. The console's certificate
is pinned: on the first request the server reads it, records its SHA-256 fingerprint and PEM in
~/.config/unifi-protect/trust.json (mode 600), and every connection afterwards verifies against
that one certificate as its own anchor. The host name check is replaced — not skipped — by a
fingerprint comparison, so addressing the console by IP is fine and a swapped certificate fails
hard.
This is what changed. The certificate is issued to unifi.local with no IP SAN, so reached by
an IP address, host name verification could never pass however the certificate was trusted; the
old advice was to set NODE_EXTRA_CA_CERTS and address the console by a resolvable name, or
give up and disable verification. Pinning removes both requirements.
Pinning is trust-on-first-use, and it is only as good as that first moment. The fingerprint is
logged when it is learned and reported by unifi_protect_auth_status; set
UNIFI_PROTECT_FINGERPRINT to make the trust explicit instead, and a console that presents
anything else is then refused rather than adopted. If the console legitimately reissues its
certificate, delete the trust file.
UNIFI_PROTECT_VERIFY_TLS=false still turns checking off entirely, scoped to this server's own
requests through an undici dispatcher — it is not NODE_TLS_REJECT_UNAUTHORIZED, so nothing else
in the process is affected. There should no longer be a reason to use it; the startup banner
prints tls=UNVERIFIED on every run when you do.
Blast radius. With the defaults, the worst an agent can do is read your cameras and write
image files into the snapshot directory. With UNIFI_PROTECT_ALLOW_WRITES=1 it can additionally
reconfigure devices, stop a camera recording, and reboot a camera or the whole console. Use a
Local-Access-Only account with View Only rights, and leave writes off unless you need them.
Configure
Variable | Required | Default | What it does |
| yes | — | Console IP or hostname. |
| yes | — | Console login |
| yes | — | Its password |
| no | — | 2FA code. Expires in ~30s — prefer |
| no |
| Verify the console's certificate. Off disables pinning entirely |
| no | — | Pin this SHA-256 up front instead of learning it on first use |
| no |
| Where the pinned certificate is remembered, mode 600 |
| no |
| Register the 12 mutating tools |
| no |
| Cached session, mode 600 |
| no |
| Where images and exports are written |
| no |
| Config file location |
| no |
| Retries on 401 / 429 / 5xx |
| no |
| Refuse a download larger than this |
| no |
| Camera id→name cache lifetime, seconds |
| no | — | Verbose request logging to stderr |
The config file mirrors these as camelCase JSON (host, username, verifyTls, …). It is
strict: an unknown key is an error rather than a silent no-op. Environment variables win over
the file, field by field, so a one-off UNIFI_PROTECT_ALLOW_WRITES=0 still beats a file that
says true.
Create an account for it
UniFi OS → Settings → Admins & Users → Add User → Local Access Only, with Protect permissions and View Only unless you plan to enable writes.
Use a local account rather than your Ubiquiti (SSO) one. Cloud accounts frequently cannot log in locally at all, and a scoped local account keeps this server away from the rest of the console.
Quick start
A. npx
UNIFI_PROTECT_HOST=192.168.1.1 UNIFI_PROTECT_USERNAME=mcp UNIFI_PROTECT_PASSWORD=… \
npx -y @mgcrea/mcp-unifi-protectB. Docker (stdio)
docker run --rm -i \
-e UNIFI_PROTECT_HOST=192.168.1.1 \
-e UNIFI_PROTECT_USERNAME=mcp \
-e UNIFI_PROTECT_PASSWORD=… \
ghcr.io/mgcrea/mcp-unifi-protectC. From source
pnpm install && pnpm build
node dist/cli.jsInspect the tools
printf '%s\n' \
'{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"cli","version":"0"}}}' \
'{"jsonrpc":"2.0","method":"notifications/initialized"}' \
'{"jsonrpc":"2.0","id":2,"method":"tools/list"}' \
| node dist/cli.js 2>/dev/null | jq -r '.result.tools[]?.name'Tools
22 read tools, plus 10 more when writes are enabled.
Tool | What it does | Writes |
| Log in and make a real call, reporting whether the console is reachable and on what Protect version | — |
| Force a fresh login; the only way to supply a 2FA code | — |
| Drop the cached session and delete the session file | confirm |
| Console model, Protect version, storage, device counts | — |
| Every camera, summarized | — |
| One camera's complete record (large) | — |
| Capture a frame now, to a file or inline | — |
| A PTZ camera's saved preset slots | — |
| A PTZ camera's saved patrol routes | — |
| Audit every camera for inconsistent or self-defeating settings | — |
| Search recorded events over any time range | — |
| One event's full detection metadata | — |
| The frame that triggered a detection | — |
| Up to 6 frames at once, inline — how you tell a person from a branch | — |
| Export footage as an MP4 on disk | — |
| Floodlights, with state and brightness | — |
| Sensors, with temperature / humidity / light readings | — |
| Viewport devices and what each displays | — |
| Chimes, volume, paired doorbells | — |
| Saved camera grid layouts | — |
| Who can sign in to Protect | — |
| Escape hatch: call any private endpoint directly | GET only unless writes |
| Name, mic, status LED, OSD overlays | ✅ |
| Which objects and sounds a camera detects — the gate below | ✅ |
|
| ✅ |
| Reboot one camera | ✅ confirm |
| Brightness, on/off, PIR sensitivity | ✅ |
| Name, which capabilities report | ✅ |
| Put a live view on a screen | ✅ |
| Volume, name | ✅ |
| Console name, timezone, global recording | ✅ |
| Reboot the console | ✅ confirm |
Resources and prompts
Three resources carry the standing facts a question needs before a tool is chosen, so a client can attach them once instead of spending a call per question:
Resource | Why it exists |
| The console's time zone, so "1am" is read as the local clock rather than UTC |
| What each camera will actually detect, what its zones ask for, and where they differ |
| Named groups of cameras, so a question about a place resolves to ids |
Two prompts carry the procedure, which is the part a tool list cannot express:
check_camera_settings— run the audit and interpret it, changing nothing. Several findings have two valid opposite fixes, and which is right depends on what the camera is for.who_passed— find who was present in a window, and fall back to motion frames on any camera whose detector is off rather than reporting a zero count as an absence. It takes the question as one free-text argument, so quote it: slash-command arguments are split shell-style and mapped positionally, sowho_passed in front of the house last night?arrives as just"in", whilewho_passed "in front of the house last night?"arrives whole. A single-word question is treated as that truncation and refused rather than answered.
Worked example: what happened at the front door last night
// 1. Which cameras are there?
{"name": "unifi_protect_list_cameras", "arguments": {}}
// → [{ "id": "661a…", "name": "Front Door", "hasSmartDetect": true,
// "smartDetectTypes": ["person","package"], "recordingMode": "detections", … }]
// 2. People seen overnight. Note the camera NAME comes back resolved.
{"name": "unifi_protect_list_events", "arguments": {
"start": "2026-08-29T22:00:00Z", "end": "2026-08-30T07:00:00Z",
"types": ["smartDetectZone"], "smartDetectTypes": ["person"]}}
// → { "count": 3, "events": [
// { "id": "9f3c1a02-…", "start": "2026-08-30T02:14:07.000Z", "camera": "Front Door",
// "smartDetectTypes": ["person"], "score": 94, "hasThumbnail": true }, … ] }
// 3. Look at the one at 02:14 — pass the event's own id.
{"name": "unifi_protect_get_event_thumbnail",
"arguments": {"eventId": "9f3c1a02-…", "output": "image"}}
// 4. Pull the footage around it.
{"name": "unifi_protect_export_video", "arguments": {
"cameraId": "661a…", "start": "2026-08-30T02:13:30Z", "end": "2026-08-30T02:15:00Z"}}
// → { "path": "/Users/you/.cache/unifi-protect/front-door-….mp4", "bytes": 18432000,
// "durationSeconds": 91.04, "startsAt": "2026-08-30T02:13:28.960Z",
// "audio": { "durationSeconds": 91.04, "sampleRate": 16000 }, … }Traps worth knowing
This wraps Protect's private API, not the official one. Ubiquiti publishes an Integration
API at /proxy/protect/integration/v1 with an OpenAPI spec and an X-API-KEY header. It is not
used here, because it has no historical query capability at all — the only query parameters
in its entire spec are channel, highQuality and qualities, and events exist solely as a
live WebSocket. "What happened last night" is unanswerable through it. The private API answers
that, at the cost of being undocumented and liable to change between Protect releases. This was
built and verified end-to-end against a live UNVR running Protect 7.2.105.
unifi_protect_get_system_info reports the version you are actually running, and
unifi_protect_request reaches any endpoint that moves.
Two shapes already changed between 6.x and 7.x, both found by running this against a real console, and both now handled in either form:
Storage moved. 6.x had
nvr.storageInfowithtotalSize/totalSpaceUsed. By 7.2 that key is gone; the numbers live undernvr.systemInfo.storageandnvr.storageStats, with per-disk health insystemInfo.ustorage.disks.A camera has no
ledLevel. The 0-6 brightness that looks like it belongs there is a floodlight field; a camera's LED is the on/offledSettings.isEnabled. Sub-objects also deep-merge on PATCH, so setting one OSD overlay preserves the others — verified by writing to a live camera and reading it back.An event's
thumbnailfield is not a thumbnail id you can use here. It readse-<eventId>and belongs to thethumbnails/<id>endpoint;events/<eventId>/thumbnail— the one this server calls — wants the bare event id. Passing the console's own value returns 404. So list results reporthasThumbnail: truerather than an id, andunifi_protect_get_event_thumbnailtakes the event'sid(though it tolerates ane-…value too).
Smart detection is gated in two places, and only one of them is obvious.
smartDetectSettings.objectTypes on the device is the master switch;
smartDetectZones[].objectTypes says what each zone asks for. A zone can ask for person while
the device list omits it, and the console then reports nothing at all — no error, no warning,
just an empty result forever. On the console this was built against, a doorbell had
zone: [person, vehicle, animal] against device: [animal], so a person search returned zero
across seven days while people walked past nightly.
Zero results are therefore never reported bare. unifi_protect_list_events cross-checks the
requested detection types against each camera's device list and returns a warnings array
saying the detector was off — the difference between "nobody was there" and "nothing was
looking". unifi_protect_check_settings finds the same misconfiguration across the whole
system, and unifi_protect_set_camera_detections fixes it, keeping the zones in step.
One limit worth knowing: the check reflects the camera's setting now, so a historical search over a period when the detector was off but has since been enabled gets no warning.
Some settings are reported on read but refused on write. smartDetectSettings.audioTypes
comes back containing smoke_cmonx, and a PATCH containing it fails with
400 The smart detection feature is not enabled for: smoke_cmonx. Any read-modify-write that
echoes the list back therefore breaks. unifi_protect_set_camera_detections filters against
featureFlags.smartDetectAudioTypes and reports what it dropped.
Camera filtering happens on the console, and the parameter must be repeated. /events
accepts cameras=<id>, repeated once per camera. A comma-separated list is accepted and
silently matches nothing. This mattered more than it looks: filtering client-side instead
fetches the newest limit events across all cameras and discards the rest, so a quiet camera
over a long window came back empty while reporting a successful search.
Times are milliseconds, and getting it wrong fails silently. The console takes JavaScript
millisecond timestamps. A Unix seconds value is not rejected — it is read as a moment in 1970,
so the query succeeds and returns an empty list, which reads as "nothing happened". Every time
argument here accepts ISO 8601, a relative expression ("2h ago", "30m", "7d") or "now",
and a ten-digit number is refused with the corrected value in the error.
Local forms are also accepted — "1am", "01:30", "2026-08-30 01:00" — and read in the
console's own time zone, because a question about last night is a question about the clock
where the cameras are. A bare time of day resolves to its most recent occurrence, and start
anchors to the window's end, so "1am to 6am" stays one coherent night however late it is asked.
Event search is always filtered by type. Omitting types entirely triggers a pagination bug
in Protect where the console ignores the window and returns the wrong slice. unifi_protect_list_events
always sends an explicit list, defaulting to motion, smart detections and rings.
Footage only exists if the camera was recording. An empty event search may mean the camera's
recording mode is never, not that nothing happened. unifi_protect_list_cameras shows the mode.
An export starts early, and not by a fixed amount. The console cuts on the keyframe before
the requested start, so the file opens up to a second or so before it and runs that much long.
unifi_protect_export_video reads the file's own duration and reports startsAt, the end
minus the duration; against overlapping exports of one camera that lands within ~50 ms.
Footage from several cameras lined up on the requested start instead is off by a different
amount per camera. Some cameras' audio clocks also run ~0.2% slow, so their audio
track is a few seconds short of the video over an hour.
Plates live on the vehicle, not on the event. On Protect 7.x, metadata.licensePlate is
absent; each tracked object sits in metadata.detectedThumbnails, with the plate as the
vehicle's name. unifi_protect_list_events reads it from there, and counts the people the
event tracked.
Snapshots are forced. Without that the console can return a cached frame minutes old, which is indistinguishable from a current one.
A cloud account may not work. Ubiquiti SSO accounts frequently cannot log in locally. Create a Local Access Only user.
Troubleshooting
The server does not appear, or shows Connection closed. It should never exit on missing
credentials — run it by hand with the same environment and read stderr. Everything it logs goes
to stderr, because stdout is the protocol channel.
Only unifi_protect_auth_status is listed. No console is configured. Call that tool; it
returns the setup steps as data.
A tool I expected is missing. The write tools are not registered unless
UNIFI_PROTECT_ALLOW_WRITES=1. That is the design, not a bug — an absent tool cannot be called,
whereas a refused one invites an agent to keep trying.
self-signed certificate errors. These should no longer happen: the certificate is pinned on the first request, which works against an IP address. If one appears, the console is presenting a different certificate than the one pinned — delete ~/.config/unifi-protect/trust.json if it was legitimately reissued, and investigate if it was not
unless you have installed a trusted certificate on the console.
Cloud mode returns 403 user cannot access host in the organization. The key is
valid but was issued in an organization that does not contain that console. Check the
console appears in curl -H "X-API-KEY: $KEY" https://api.ui.com/v1/hosts; if the web
dashboard shows it but that call does not, they are different organizations.
I set an API key for local mode and everything returns 401. Local mode cannot use an
API key — see Why local mode needs a username and password.
Set UNIFI_PROTECT_USERNAME and UNIFI_PROTECT_PASSWORD, or switch to cloud mode,
where a key is all you need.
A local API key returns 401 on everything. API keys are per-console. A key created
on your Network gateway is not valid on the NVR running Protect — and the NVR rejects an
unknown key with exactly the same 401 it gives a fabricated one, so the message cannot
distinguish "wrong console" from "wrong key". Create the key on the console you are
addressing, or use cloud mode.
Everything returns 401. Check the account is a local one, and that it has Protect
permissions. unifi_protect_auth_status distinguishes "cannot log in" from "logged in but
forbidden".
A tool that used to work now returns 404. Compare the Protect version from
unifi_protect_get_system_info against 7.2.105 above; an upgrade may have moved the endpoint.
unifi_protect_request is the workaround while it is fixed — it reaches any path under
/proxy/protect/api directly, which is how both of the 6.x→7.x changes above were pinned down.
Storage shows as nearly full. That is normal on an NVR: isRecycling: true means the
console continuously overwrites the oldest footage rather than stopping. get_system_info says
so inline so it does not read as a fault.
What has been verified against real hardware
Cloud mode was verified end-to-end against a live console over the Site Manager
connector: auth_status reachable, camera list, and event search returning real
detections with their camera names resolved — all authenticated by an API key alone,
with no local account anywhere in the picture.
Local mode was built and exercised end-to-end against a live UNVR4 on Protect 7.2.105 with 12 cameras, one floodlight and one chime. Every read tool was run; the write tools were exercised with no-op writes — each value set to the value it already held — and the device state read back unchanged afterwards. That run is also what caught three bugs this README's earlier drafts described wrongly: the storage layout, the event-thumbnail id, and two camera fields that do not exist.
Certificate pinning was verified against that same console on 2026-08-31, addressed by IP
(192.168.6.3) with verification ON — the case that was impossible before. Confirmed in one run:
the certificate is captured and its fingerprint logged on first contact; the trust file is written
mode 600; a restart reuses the pin without re-capturing; a correct UNIFI_PROTECT_FINGERPRINT
(colons and all) is accepted; and a wrong one is refused with a message naming both fingerprints,
surfacing as a failed tool call rather than a server that never starts. auth_status reported
reachable: true on Protect 7.2.105 throughout, and a session cached by the previous release was
restored unchanged — the on-disk format did not move.
Two tools remain unverified for want of hardware, and are marked here rather than left to look tested:
unifi_protect_update_sensor— no UP Sense device on the test console (sensorsis empty). Note that a Protect floodlight has its own built-in PIR, reported asisPirMotionDetectedand tuned viapirSensitivity; that is part of the light, so it isunifi_protect_update_lightthat controls it, not this tool. A garden lamp is not a sensor device.unifi_protect_update_viewer— no Viewport device on the test console.
Both follow the same PATCH shape as the tools that were verified, so they are likely correct, but "likely" is the honest word until someone runs them.
Not implemented
The realtime WebSocket at /proxy/protect/ws/updates is not wired up here, though
@mgcrea/unifi-protect now implements it — connectEventStream and ProtectStore are a
decoded frame stream and a live device cache, and this server deliberately uses neither.
The reason is that they are the wrong shape for an MCP server. Both want a long-lived process
that bootstraps at start-up and stays subscribed; this server must start with no credentials and
no connectivity at all, answering unifi_protect_auth_status until it is configured. It would
also gain little: because this wraps the private API, event history is already available over
REST through unifi_protect_list_events, which is what a subscription would have been for.
Adopting it would mean deferring the bootstrap until credentials appear — worth doing if a tool
ever needs push rather than poll, and not before.
Develop
pnpm install
pnpm lint && pnpm format:check && pnpm typecheck && pnpm test && pnpm buildPublish:
pnpm release minor # bump, commit, tag (patch|minor|major)
git push --follow-tags # CI publishes to npm + GHCR from the tagLicense
MIT
Available Tools
22 toolsunifi_protect_auth_loginUniFi Protect: Auth LoginA
Force a fresh login to the console, replacing any cached session. Normally unnecessary — the server logs in on demand and re-authenticates automatically on a 401. Use it to supply a two-factor code, which cannot be done unattended: the code is single-use and expires in about 30 seconds, so it is passed here once and the resulting session is then cached and reused.
Inputs: totp.
unifi_protect_auth_logoutUniFi Protect: Auth LogoutADestructiveIdempotent
Drop the cached session and delete the session file. The next call logs in again from the configured username and password, so this does not lock anything out — use it to clear a session after changing accounts, or to remove the cookie from disk.
Inputs: confirm (required).
unifi_protect_auth_statusUniFi Protect: Auth StatusARead-only
Check whether this server can actually reach your UniFi Protect console. By default it logs in and makes a real call, so the answer reflects the console rather than cached state — this is the tool to run when something is not working. Reports the host, the account, the Protect version, whether TLS is verified, and whether writes are enabled; when nothing is configured it returns the exact setup steps instead. Call this first when a tool you expected is not listed: an absent tool means missing configuration or writes being off, not a bug.
Inputs: probe.
unifi_protect_check_settingsUniFi Protect: Check SettingsARead-only
Check every camera and the console for settings that are inconsistent, or that mean the system is not doing what someone believes it is. Finds detectors that look enabled but are gated off, cameras keeping no footage, offline devices, motion detection switched off, and storage about to stop recording. This is the tool for "are my camera settings correct" — the checks encode traps that are invisible in the Protect UI, notably a detection zone asking for an object type the device list blocks. Read-only: it reports findings and names the tool that would fix each one, and changes nothing.
Inputs: no named parameters.
unifi_protect_export_videoUniFi Protect: Export VideoARead-only
Export recorded footage from one camera over a time range as an MP4 file on disk. Always writes to a file and returns the path — video is never returned inline. Size grows quickly with the window: expect tens of megabytes per minute at full quality, and the call fails rather than exhausting memory if the export exceeds UNIFI_PROTECT_MAX_DOWNLOAD_BYTES. Footage only exists if the camera was recording at the time, so check the recording mode before concluding that nothing happened. The file starts on the keyframe before start, so it runs a little long at the front: startsAt is when its first frame was recorded, which is what to line exports from several cameras up on. audio gives the sound track's length and sample rate, or null when the export has no sound.
Inputs: end (required), start (required), cameraId (required), channel, +1 more.
unifi_protect_get_cameraUniFi Protect: Get CameraARead-only
Get one camera's complete record — every setting the console holds, including encoder channels, motion and smart-detection zones, privacy masks, OSD and LED settings, ISP tuning and live statistics. This is large (roughly 8-15 KB of JSON). Prefer unifi_protect_list_cameras unless you specifically need a field it does not carry.
Inputs: cameraId (required).
unifi_protect_get_camera_snapshotUniFi Protect: Get Camera SnapshotARead-only
Capture a still frame from a camera as it looks right now. Writes the JPEG to disk and returns its path, size and content type by default. Set output="image" to get the frame inline instead so a vision model can actually look at it — that costs roughly 300,000 to 700,000 characters of context per call, so choose it deliberately rather than by default. A fresh capture is forced; without that the console can hand back a cached frame that is minutes old.
Inputs: cameraId (required), output, savePath, highQuality.
unifi_protect_get_eventUniFi Protect: Get EventARead-only
Get one event's full record, including detection metadata the search results leave out — per-object tracking, detected zones, licence plate text and vehicle attributes where the camera captured them. Use the id from unifi_protect_list_events.
Inputs: eventId (required).
unifi_protect_get_event_thumbnailUniFi Protect: Get Event ThumbnailARead-only
Fetch the still image Protect captured for an event — the frame that triggered the detection. Writes it to disk and returns the path by default; set output="image" to return it inline for a vision model to look at, which costs a large amount of context. Pass the event's id from unifi_protect_list_events; results showing hasThumbnail: true have one.
Inputs: eventId (required), output, savePath.
unifi_protect_get_event_thumbnailsUniFi Protect: Get Event ThumbnailsARead-only
Fetch the still frames for SEVERAL events at once and return them inline to look at. This is the tool for answering who or what was actually there, and it matters most when a camera has no smart detection: motion events carry no classification, so the only way to tell a person from a branch is to look. Prefer this over calling unifi_protect_get_event_thumbnail repeatedly. Costs roughly 1-2K tokens per image, so it is capped at 6 — pick the events worth seeing from unifi_protect_list_events rather than sweeping a whole night. Events that have no thumbnail are reported by id instead of failing the call.
Inputs: eventIds (required), output.
unifi_protect_get_system_infoUniFi Protect: Get System InfoARead-only
Overview of the console: model, Protect version, firmware, timezone, uptime, storage use and how many devices of each type are adopted. Worth calling first on an unfamiliar system. The reported Protect version matters: this server talks to Protect's private API, which Ubiquiti changes between releases, so a version that differs from the one in the README is the first thing to check if a tool starts returning 404.
Inputs: no named parameters.
unifi_protect_list_camerasUniFi Protect: List CamerasARead-only
List every camera on the console with its id, name, connection state, recording mode, firmware and what it can do (PTZ, package camera, smart detection, and which object types it detects), and whether its recordings carry sound (recordsAudio: a mic, switched on, with audio kept in the recording). Returns a summary rather than the console's full camera record, which runs to thousands of fields across encoder profiles, zones and feature flags — use unifi_protect_get_camera when you need all of it for one camera.
Inputs: no named parameters.
unifi_protect_list_chimesUniFi Protect: List ChimesARead-only
List UniFi Protect chimes, their volume, and which doorbell cameras each is paired to.
Inputs: no named parameters.
unifi_protect_list_eventsUniFi Protect: List EventsARead-only
Search recorded events over any time range — motion, smart detections (person, vehicle, animal, package, licence plate), doorbell rings, and camera connection changes. This is the tool for questions like "what happened at the front door last night". Each result carries its camera's NAME as well as its id, so no second lookup is needed. Smart detections also say what was seen: people (how many people that one event tracked) and vehicles (each with its licence plate, colour and type where read). Times may be given in the console's own local clock ("1am"), which is what a question about last night means. READ ANY warnings IN THE RESULT BEFORE REPORTING A COUNT: a camera with the detector switched off returns zero matches, which is not the same as nothing having happened, and this tool says which case it is. Narrow with types, smartDetectTypes and cameraIds wherever you can: a busy system logs thousands of motion events a day.
Inputs: end, limit, order, start, +5 more.
unifi_protect_list_lightsUniFi Protect: List LightsARead-only
List UniFi Protect floodlights with their connection state, whether the light is currently on, whether PIR motion is being detected, and brightness.
Inputs: no named parameters.
unifi_protect_list_liveviewsUniFi Protect: List LiveviewsARead-only
List the saved live views — the named camera grid layouts shown on viewers and in the Protect app. The returned id is what unifi_protect_update_viewer needs to put a layout on a screen.
Inputs: no named parameters.
unifi_protect_list_ptz_patrolsUniFi Protect: List PTZ PatrolsARead-only
List a PTZ camera's saved patrol routes. See unifi_protect_list_ptz_presets for why there is no tool to start or stop one.
Inputs: cameraId (required).
unifi_protect_list_ptz_presetsUniFi Protect: List PTZ PresetsARead-only
List a PTZ camera's saved preset positions, with the slot number each one lives at. Only meaningful for cameras reporting hasPtz: true in unifi_protect_list_cameras. There is no tool to MOVE a PTZ camera or run a patrol: those commands exist only on Ubiquiti's official Integration API (a separate X-API-KEY auth this server does not use), not on the private API this server wraps — presets are created and driven from the Protect app itself.
Inputs: cameraId (required).
unifi_protect_list_sensorsUniFi Protect: List SensorsARead-only
List UniFi Protect sensors with their current readings — temperature, humidity, light level — plus open/closed state, motion, and battery percentage. The readings are lifted out of the console's per-metric history arrays, which are far larger than the values themselves.
Inputs: no named parameters.
unifi_protect_list_usersUniFi Protect: List UsersARead-only
List the accounts that can sign in to Protect, with their role and last login. Useful for auditing who has access to the cameras.
Inputs: no named parameters.
unifi_protect_list_viewersUniFi Protect: List ViewersARead-only
List UniFi Protect Viewport devices and which live view each is currently displaying.
Inputs: no named parameters.
unifi_protect_requestUniFi Protect: RequestARead-only
Escape hatch: call any private Protect API endpoint directly, relative to /proxy/protect/api. This exists because the private API is undocumented and Ubiquiti moves endpoints between Protect releases — when a wrapped tool starts returning 404, this reaches the replacement without waiting for a new version of this server. Responses are returned RAW and unshaped, so a broad endpoint like bootstrap can return hundreds of kilobytes; prefer the wrapped tools, which summarize. Writes are DISABLED: only GET is permitted. Set UNIFI_PROTECT_ALLOW_WRITES=1 to allow mutations.
Inputs: path (required), body, query, method.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 22 tools
Nearly every tool targets a distinct resource+action, and the descriptions actively disambiguate the close pairs (list_cameras vs get_camera, get_event_thumbnail vs get_event_thumbnails, auth_status vs auth_login vs auth_logout). The main residual risk is the singular/plural thumbnail pair and the three-tool auth cluster, which an agent could still pick wrongly under time pressure, but descriptions largely resolve this.
Every tool uses the same unifi_protect_ prefix followed by a consistent snake_case verb_noun pattern (list_cameras, get_event, export_video, check_settings). No mixing of conventions or casing anywhere in the set.
22 tools is at the upper edge, but the domain genuinely spans several device classes (cameras, sensors, chimes, lights, viewers), events, media export, and auth, so most tools earn their slot. It is slightly heavy and could be trimmed (e.g. collapsing the PTZ preset/patrol pair or the thumbnail pair), but not bloated.
Read/monitoring coverage is excellent and explicit gaps (PTZ movement, patrol control) are documented as upstream API limits. However, the surface is essentially read-only — there is no way to change camera settings, arm/disarm, trigger a chime, or update anything, despite check_settings naming the fixes it cannot perform, which is a notable lifecycle gap for a management server.
Maintenance
Related MCP Connectors
Ask your security cameras anything and set up alert rules in a sentence. Built into Agent DVR.
Watch cameras and video with AI vision and get alerts in Slack, Discord, Telegram or a webhook.
Give the physical world an API: cameras become typed endpoints your agent can read and act on.
Control your Tesla from your AI assistant - climate, charging, access, and security.
Related MCP Servers
- AlicenseDqualityDmaintenanceEnables comprehensive management of UniFi network infrastructure through the UniFi Cloud API, including device control, client management, camera settings, and access door control through natural language.39MIT
- AlicenseBqualityDmaintenanceEnables AI assistants to manage and monitor UniFi Network Controllers through natural language. Provides 25 read-only tools for discovering devices and clients, viewing security configurations, analyzing network statistics, and exporting configuration data.41MIT
- AlicenseBqualityDmaintenanceEnables AI assistants to manage UniFi network infrastructure through 50+ tools covering devices, clients, networks, WiFi, firewall rules, and guest access using the official UniFi Network API.525MIT
- AlicenseBqualityBmaintenanceExposes UniFi Protect's Integration REST API as tools for Claude Code and other MCP clients, providing 73 tools to manage cameras, lights, sensors, and more with safety features like read-only mode and dry-run support.38240 npm1MIT