The CLI in scripts and servers, through MEMOKET_API_TOKEN
Yes
Webhook subscriptions created with the CLI
No
Webhook subscriptions created with the API
Yes
Create a key
In Memoket for desktop
Keys live on the Personal API keys page in Memoket for Mac or Windows. If you do not have it yet, download Memoket for desktop.
1
Open Connectors
In Memoket for Mac or Windows, click Connectors in the sidebar.
2
Open Personal API keys
In the API section, click Personal API keys.
3
Create the key
Click Create key, enter a name, and choose when it expires: 30 days, 90 days, or 1 year (1 year by default).
4
Copy it
Copy the key and store it somewhere safe. It is shown only once.
With the CLI
Sign in once, then create a key. Without --scope, the key gets every read permission, the same as a key from the desktop app. To limit it, pass --scope with a comma-separated list of recordings:read, summaries:read, transcripts:read, and search:read, for example --scope summaries:read,transcripts:read. --expires-in sets the lifetime in days; without it, the key lasts 1 year.
Terminal
memoket login
memoket token create --name "My script"
Keys from the desktop app and the CLI are one list. A key created in one place shows up in the other.
Manage keys in Memoket for desktop
The Personal API keys page lists every key you created in Memoket for desktop or with the CLI that has not been revoked, newest first.
Name
The name you gave the key, tagged Desktop or CLI for where it was created. A key created with the CLI without a name is named Personal key 1, Personal key 2, and so on.
Key
The last four characters, such as ••••3f9a. The full key is shown only once, when you create it.
Last used
The day the key was last used, or Never used.
Expires
The expiry date. An expired key shows it in red.
Actions
Revoke for an active key. Delete removes an expired one from the list.
Use a key
Send it in the Authorization header of every API request.
Header
Authorization: Bearer mtok_live_...
For curl and for the CLI in a script, set it as MEMOKET_API_TOKEN. For an AI app that asks for a token, paste the full key into its token or secret field.
Shell
export MEMOKET_API_TOKEN=mtok_live_...
Key details
Format
Starts with mtok_live_.
Access
Keys from the desktop app, and CLI keys created without --scope, can read recordings, transcripts, briefs, and summaries, get audio download links, and manage webhook subscriptions. A CLI key created with --scope has only the permissions you list.
Expiry
Set when the key is created: 30 days, 90 days, or 1 year in the desktop app, or --expires-in in the CLI. Without --expires-in, a CLI key lasts 1 year. A key does not expire from not being used.
Listed in
Personal API keys in Memoket for desktop and memoket token list.
Limits
Up to 10 active keys per account. Revoked and expired keys do not count toward this. You can create up to 10 keys a minute, rotations included.
Revoke or rotate a key
Task
Desktop app
CLI
Revoke
Click Revoke next to the key.
memoket token revoke <id>
Rotate
Not available
memoket token rotate <id>
A revoked key stops working right away. Requests that use it return 401. Rotating creates a new key with the same name and revokes the old one.
Treat a key like a password. Keep it out of code, prompts, and chat, and store it in a secret manager.
Some apps cannot open a browser to sign in. Instead they ask for a server URL plus a token, an API key, a secret, or a header. Connect these apps with an API key.
1
Create an API key
Create one in Memoket for desktop or with the CLI, as shown on API keys. Copy it right away; it is shown only once.
2
Add the server
Enter https://mcp.memoket.ai/mcp as the server URL. If the app asks for a transport, choose Streamable HTTP.
3
Add the key
If the app has a field named token, API key, or secret, paste the key itself. If it asks for a header, add Authorization with the value Bearer mtok_live_....
Add MCP server
Namememoket
Server URLhttps://mcp.memoket.ai/mcp
TransportStreamable HTTP
HeaderAuthorization: Bearer mtok_live_••••••
CancelSave
What a typical app asks for. Field names differ from app to app.
Apps that use a config file take the key as a header:
Tool calls from the CLI are limited to 120 a minute and 3,000 an hour per account. See Rate limits.
Get an audio download link
Gets a download link for the original audio of one recording. Use the id from list_conversations, or data.conversation_id from a webhook event. One recording per command.
Terminal
memoket audio download 7Q2KX4M3ZJ5WA
The CLI prints the link with the recording's title, file type, and size. Open the link in a browser, or save the file with any download tool:
Terminal
curl -L -o recording.opus "<link>"
Link
Works for 5 minutes, so start the download right away. Run the command again for a new link.
Format
The original audio, not converted. Recordings from a Memoket device are .opus. Phone, desktop, imported, and merged recordings are .m4a.
Length
Any length.
Permission
Needs recordings:read. memoket login, keys from the desktop app, and CLI keys created without --scope have it.
In scripts, memoket call download_audio conversation_id=<id> returns the same link as JSON, as described in Get an audio download link.
Use with an agent
Any agent that can run terminal commands can use Memoket through the CLI, for example:
Paste this prompt into your agent to set it up, then ask it to use Memoket.
Agent prompt
Install the official Memoket CLI here and prove that one read-only call works.
1. Check node --version. Memoket needs Node.js 18 or newer.
Stop and tell me if it is missing.
2. Run npm install -g @memoket-ai/cli, then memoket --version.
3. Run memoket login. When the browser asks me to sign in, wait for me.
4. Run memoket doctor, then memoket call list_conversations count=1.
5. Report the CLI version and whether the call worked.
Do not print recording content or any key.
Exit codes
Code
Meaning
0
The command worked.
1
The command failed or stopped. The message says why.
Troubleshooting
Message or problem
What to do
command not found: memoket
Open a new terminal. If it still fails, check that npm's global folder is in your PATH.
The browser does not open
Where no browser can open, create a key in Memoket for desktop and set MEMOKET_API_TOKEN. See API keys.
not logged in
Run memoket login, or set MEMOKET_API_TOKEN.
token lacks recordings:read
The key was created with --scope. Create a new one without it.
recording not found
The ID is wrong, belongs to another account, or the recording was deleted for good.
The audio is not ready yet
The recording is still uploading. Try again in a few minutes.
A key was lost before you saved it
Revoke it with memoket token revoke <id> and create a new one.
Read your recordings, transcripts, briefs, and summaries over HTTPS with an API key.
Base URL
Base URL
https://api.memoket.ai
Authentication
Send an API key in the Authorization header. To create one, see API keys. The samples on these pages read the key from MEMOKET_API_TOKEN.
Header
Authorization: Bearer mtok_live_...
Shell
export MEMOKET_API_TOKEN=mtok_live_...
Call a tool
The API exposes Memoket's read tools. Every tool works the same way: send POST /v1/tools/{name} with a JSON body, and the response is the tool's result.
A link to download the original audio of one recording. The link works for 5 minutes.
recordings:read
Keys from the desktop app, and CLI keys created without --scope, have every permission listed here.
Search is not part of the API. Calling search_recordings returns 404. To find recordings, list them by time with list_conversations and start_time, or search with the CLI or an AI app.
IDs and pages
List recordings
id: 7Q2KX4M3ZJ5WA
id
Get recordingsget_conversations
Get a transcriptget_transcripts
Get what participants saidget_transcripts_by_participant
Get summariesget_summaries
Get briefsget_brief
Get an audio download linkdownload_audio
Find a recording first, then pass its id to the endpoint that reads it.
Recording IDs are opaque strings such as 7Q2KX4M3ZJ5WA. Store and pass them exactly as returned.
The same recording ID is id in list_conversations and get_conversations results, conversation_id in get_transcripts and download_audio, and data.conversation_id in webhook events. get_conversations, get_transcripts_by_participant, get_summaries, and get_brief take it as an entry of conversation_ids.
Participant IDs for get_transcripts_by_participant come from participant_details in list_conversations or get_conversations.
list_conversations returns 100 recordings per page by default; set count to get up to 200. Pass the returned next_offset as offset to get the next page.
get_transcripts returns up to 100 lines or 4,000 characters per page. Pass next_offset and next_text_offset back to continue.
Asking a read tool for a recording from another account returns an empty result, not an error. download_audio returns 404 instead.
Rate limits
Tool calls are limited per account. The API, the CLI, and MCP each have their own limit, and every key and sign-in on your account shares it, so more keys do not raise it.
Per minute
Per hour
API
60
1,200
CLI
120
3,000
MCP
120
1,200
The per-minute and per-hour limits apply at the same time. Over a limit, the API returns 429 with a Retry-After header that says how many seconds to wait. Limits on creating keys are on API keys.
Get an HTTPS request when a recording is uploaded or a summary is ready.
Events
Event
Sent when
recording.uploaded
A recording finishes uploading. Sent once per recording, before the transcript or summary is ready. Includes a link to download the original audio, valid for 2 hours.
summary.completed
A recording's summary is ready. Sent again each time a new summary is generated for that recording.
Upload finishedrecording.uploaded
Transcript and summary are created
Summary readysummary.completed
One recording, in order.
webhook.ping is sent only when you test a subscription. You cannot subscribe to it.
Payload
Every event has the same outer fields. data.conversation_id is always present. Other fields in data can be missing, so check them before use.
Event ID, prefixed evt_. The same on every retry, so use it to skip duplicates.
type
The event name.
created_at
When the event happened.
data.conversation_id
The recording ID, the same one the tools and memoket audio download use.
data.download_url
In recording.uploaded: a link to the original audio file. It works for 2 hours. For a new link, call download_audio with the same ID.
data.summaries[].format
markdown or html.
data.transcript.truncated
When true, get the full transcript with get_transcripts.
Create a subscription
Use your final public HTTPS URL. Localhost, private addresses, and plain HTTP are not accepted, and redirects are not followed. Each account can have up to five subscriptions, including disabled ones.
Creating, updating, and sampling subscriptions with an API key needs summaries:read and transcripts:read. Keys from the desktop app have both.
Verify signatures
Each request carries these headers. Compute an HMAC-SHA256 of the timestamp, a dot, and the exact raw body, using the whole secret including whsec_ as the key. Compare it in constant time, and reject timestamps older than five minutes.
Header
Value
X-Memoket-Signature
sha256= followed by the hex digest.
X-Memoket-Timestamp
Unix time in seconds when this attempt was signed.
import hashlib, hmac, time
def verify(secret: str, headers, raw_body: bytes) -> bool:
h = {k.lower(): v for k, v in headers.items()}
timestamp = h.get("x-memoket-timestamp", "")
received = h.get("x-memoket-signature", "")
if not timestamp.isdigit() or abs(time.time() - int(timestamp)) > 300:
return False
digest = hmac.new(secret.encode(), timestamp.encode() + b"." + raw_body,
hashlib.sha256).hexdigest()
return hmac.compare_digest("sha256=" + digest, received)
Verify the raw body. Parsing and re-serializing the JSON changes the bytes and breaks the check.
Delivery and retries
Delivery
+1 min
+5 min
+15 min
+15 min
+15 min
After 20 failed attempts in a row, retries included, the subscription is disabled.
A failed delivery is retried up to 5 times, about 1, 5, 15, 15, and 15 minutes apart.
Return a 2xx status within 5 seconds. Do longer work after you reply.
A timeout, an error status, or a redirect counts as a failure.
The same event can arrive more than once. Skip events whose id you have already handled.
A disabled subscription gets no events. After you fix your endpoint, delete the subscription and create it again. The new subscription has a new secret.
Manage subscriptions
Task
API
CLI
List
GET /v1/webhooks
memoket webhook list
Change URL or events
PATCH /v1/webhooks/{id}
memoket webhook update <id>
Delete
DELETE /v1/webhooks/{id}
memoket webhook delete <id>
Send a test event
POST /v1/webhooks/{id}/test
memoket webhook test <id>
See deliveries
GET /v1/webhooks/{id}/deliveries
memoket webhook deliveries <id>
Get sample payloads
GET /v1/webhook-events/samples
Not available
{id} and <id> are the endpoint_id. PATCH takes the same url and events fields as create; send only the ones you change. The samples call takes event, one of the two event names, and an optional limit.
memoket webhook delete asks you to confirm; add --yes in scripts and agents. Test events are webhook.ping. Handle them separately so a test never runs your real workflow.
For security reasons, we cannot unsubscribe you directly from this page. Your subscription preferences will be cleared locally. To fully unsubscribe from our mailing list, please click the 'Unsubscribe' link at the bottom of any email we've sent you.
Congratulations 🎉
You are now part of the Memoket Community!
What's Next:
Check your inbox
*Our email might have entered your promotion/spam folder, please check and move our email to your Primary inbox