Parley User Guide

Parley records your meetings, transcribes them live and writes the summary, the decisions and the action items, all on your own computer. No account, no telemetry, and no cloud unless you choose to add one.

Start here

• Installing Parley: one command, no root needed.
• First run: the setup and a tour of the app.
• Recording a meeting: capturing both sides of any call.

What Parley does

• Records both sides of any call, from your microphone and your computer’s audio, so it works with every meeting app and no bot joins your call.
• Transcribes live on your computer with Whisper, on your GPU if you have one.
• Knows who said what, and learns to recognise people across meetings.
• Writes summaries from templates, with chapters, key decisions and action items.
• Answers questions about one meeting or all of them, citing where each answer came from.
• Works with your calendar to name meetings, remind you and prepare briefs.
• Exports to Markdown, PDF, Word, JSON, subtitles or your notes app.

Parley is free and open source under the MIT license. The source is on GitHub, and the website is meetparley.app.

Installing Parley

Parley runs on 64-bit (x86_64) Linux. It is distributed as an AppImage, a single file that runs on most distributions without being installed system-wide.

Requirements

• glibc 2.39 or newer. Ubuntu 24.04, Debian 13, Fedora 40, Arch, or anything newer.
• PipeWire for audio. It is the default sound server on all of the distributions above.
• FUSE (fusermount), which AppImages use to mount themselves. Most desktops already have it; if not, install your distribution’s fuse3 package.
• About 1.5 GB of disk space for the speech and summary models downloaded on first run, plus space for your recordings.

A GPU is optional. With an NVIDIA, AMD or Intel GPU and a working Vulkan driver, transcription runs on the GPU; otherwise it runs on the CPU.

Install with one command

curl -fsSL https://meetparley.app/install.sh | sh

The installer:

1. Checks that your system meets the requirements above.
2. Downloads the latest release from GitHub. It picks the Vulkan (GPU) build if it finds an NVIDIA, AMD or Intel GPU with a Vulkan driver, and the CPU build otherwise.
3. Verifies the download’s SHA-256 checksum.
4. Installs it as ~/.local/bin/parley, with the MCP server beside it as ~/.local/bin/parley-mcp, and adds Parley, with its icon, to your application menu.

No root access is needed. You can read the script before running it.

If ~/.local/bin is not on your PATH, the installer tells you. Parley still appears in your application menu either way.

Installer options

Set these environment variables on the sh side of the pipe:

For example, to force the CPU build:

curl -fsSL https://meetparley.app/install.sh | PARLEY_VARIANT=cpu sh

Install by hand

1. Download an AppImage from the latest release:

   • Parley-<version>-vulkan-x86_64.AppImage runs Whisper on your GPU. Recommended if you have one.
   • Parley-<version>-x86_64.AppImage is CPU only and runs anywhere.

2. Optionally check it against the matching .sha256 file: sha256sum -c Parley-*.AppImage.sha256

3. Make it executable and run it:

   chmod +x Parley-*.AppImage
   ./Parley-*.AppImage
   

If FUSE isn’t available and you can’t install it, run the AppImage with --appimage-extract-and-run.

Updating

Run the install command again; it replaces the old version and keeps your meetings and settings. To find out whether there is a new version, use Check for updates in the tray menu. Parley never updates itself.

Uninstalling

curl -fsSL https://meetparley.app/install.sh | sh -s -- --uninstall

This removes Parley, parley-mcp, the menu entry and the icon. Your meetings, recordings, models and settings are left in place; see Your data to remove those too.

Building from source

For CUDA builds, or to hack on Parley, see Building from source in the repository.

First run

The first time you open Parley, a short setup walks you through three steps.

1. Welcome. Click Get started.

2. Setting things up. Parley creates its local database. Click Continue once it says Ready.

3. Getting things ready. Parley downloads two models:

   • the transcription engine, Whisper large-v3-turbo (about 547 MB), and
   • the summary engine, Gemma 3 1B, the built-in summary model (about 1 GB).

   You can click Continue as soon as the transcription engine has finished. The summary model keeps downloading in the background, and summaries become available when it’s done.

   Shortly afterwards Parley also fetches two small models in the background: one that recognises speakers (about 28 MB) and one for semantic search (about 146 MB).

That’s it: Parley opens on the Record page, ready for your first meeting. Nothing you record leaves your computer unless you later choose a cloud AI provider.

Tip

The models come from Hugging Face. If a download fails, check your connection and try again; you can also manage models later under Settings → Transcription and Summary.

Finding your way around

The sidebar on the left holds everything:

• Search meetings… filters your meetings by title as you type, and after a moment also searches inside transcripts (shown under In transcripts; click one to open the meeting at that line).
• Meetings
  • Record: start and control a recording. See Recording a meeting.
  • Ask Parley: ask questions across all your meetings. See Ask Parley.
  • Import audio: turn an existing audio file into a meeting. See Importing audio.
  • Settings
• Tools
  • Action items: every action item from every meeting. See Action items.
  • Speakers: the people Parley has learned to recognise. See Speakers.
• Upcoming: your next calendar events, once you’ve added a calendar.
• Your meetings, grouped into Today, Yesterday, This week and Older. Click one to open it; see Meeting notes.

Parley also puts an icon in your system tray. See The tray icon.

Recording a meeting

Parley records two things at once: your microphone and your computer’s system audio (whatever comes out of your speakers or headphones). That captures both sides of any call, in Zoom, Teams, Meet, Slack, Discord or anything else, without a bot joining the meeting.

Starting a recording

1. Open Record in the sidebar.
2. Optionally give the meeting a title. Otherwise it is called Meeting plus the date and time.
3. Click Start recording.

Important

Recording people without their knowledge may be illegal where you live. Let everyone in the meeting know you are recording.

While recording you can Pause, Resume and Stop. The timer doesn’t count paused time.

If your calendar has an event happening now (from 5 minutes before it starts until 15 minutes after it ends), Parley names the meeting after it and links the two. A message tells you when this happens.

Choosing devices and checking levels

Click Audio to open the device panel. Choose a Microphone and a System audio source, and watch the Mic and System level bars to check that sound is coming through. Your choice is remembered for next time. You can only change devices while not recording.

The panel is closed by default because the level meters keep your microphone open, which lights up your desktop’s “microphone in use” indicator.

Default devices can also be set under Settings → Recording → Devices.

The live transcript

Speech appears in the transcript a few seconds after it is spoken. Lines from your microphone are labelled Me and lines from system audio Speaker, until Parley works out who is speaking. Text that is still being transcribed is shown in italics underneath.

The transcription runs on your computer with Whisper. Your microphone gets echo cancellation, noise suppression and automatic gain, so you don’t need headphones. The microphone and system audio are kept on separate channels in the recording.

Bookmarks

Mark an important moment with the Bookmark button, the tray menu’s Bookmark this moment, or Ctrl+Shift+B while a Parley window is focused. Bookmarks show as time chips while recording, and afterwards in the meeting’s Chapters strip. The summary gives bookmarked moments extra weight.

Live action items (beta)

Turn on Settings → Recording → Live action items to have Parley pick out action items while the meeting is still going. They appear in an Action items card on the Record page.

When you stop

Parley opens the new meeting straight away, and finishes up in the background:

1. It works out who spoke when (see Speakers).
2. If you have downloaded a larger Whisper model (large-v3), it re-transcribes the meeting with it for better accuracy. While this runs, the Record page says Refining the last meeting….
3. If Settings → Summary → Auto-summarize after recording is on, it writes the summary, and with Auto-generate chapters after recording on, the chapters. Otherwise you generate them yourself from the meeting page.

Automatic call detection

Parley can notice when a call starts. It recognises Microsoft Teams, Zoom, Slack, Discord, Webex and Jitsi Meet, and web calls in Chrome, Chromium, Firefox, Brave and Edge, by watching which applications are using your microphone.

Choose what happens under Settings → Recording → Meeting detection → When a call starts:

• Do nothing.
• Notify me (the default). A notification asks whether to record, with Record and Open Parley buttons. The tray menu also offers Record … call.
• Record automatically. Recording starts by itself and a notification reminds you to tell the others.

Turn on Stop recording when the call ends to stop automatically about 8 seconds after the call’s microphone use ends.

Note

A browser using your microphone for anything counts as a web call.

The tray icon

Parley’s tray icon turns into a red dot while recording. Its menu shows the current status and lets you:

• Start, Pause, Resume or Stop recording, and Bookmark this moment. Starting from the tray uses your default devices.
• Open Parley, open Settings or Check for updates.
• Quit.

Closing the Parley window doesn’t stop a recording: the window hides and Parley keeps running in the tray. Use Open Parley to bring it back. If you quit while recording, Parley stops and saves the recording first.

If Parley closes unexpectedly

Audio is saved in 30-second pieces while you record. If Parley crashes or your computer loses power mid-meeting, the next launch shows interrupted meetings found. For each one you can Preview the transcript captured so far, Recover it (the audio is stitched back together), or Delete it.

Where recordings are saved

Each meeting gets its own folder in ~/Documents/parley-recordings/, containing audio.mp4 (AAC audio, microphone and system audio on separate channels) plus the transcript as JSON. Settings → Recording → Open folder opens it.

If you turn off Auto-save recordings, Parley keeps the transcript but not the audio, so you won’t be able to re-transcribe those meetings later.

Meeting notes

Click a meeting in the sidebar to open it. The transcript is on the left; on the right are three tabs: Summary, Ask and Insights.

The buttons in the header let you:

• Rename the meeting (the pencil).
• Retranscribe audio; see Re-transcribing a meeting.
• Re-detect speakers; see Re-detecting speakers.
• Copy meeting as Markdown or as JSON, to paste elsewhere.
• Export meeting…; see Exporting.
• Delete meeting. This removes the transcript, summary, notes and action items from Parley. The audio file stays in your recordings folder until you delete it there.

Transcript

Each line shows who spoke, when, and what they said.

• Click ▶ to play that line’s audio.
• Click a speaker name to say who it really is; see Speakers.
• Copy transcript copies the whole transcript with timestamps.

To find something inside transcripts, use Search meetings… at the top of the sidebar, which searches every meeting. Click a result under In transcripts to open the meeting at the matching line.

The transcript text itself can’t be edited. If it’s poor, try re-transcribing with a larger model or a fixed language, and add tricky words to your custom vocabulary.

Chapters and bookmarks

The Chapters strip above the transcript lets you jump around the meeting by topic. Click Generate chapters to have the AI split the meeting into chapters (this happens automatically if Auto-generate chapters after recording is on). Chapter headings also appear inside the transcript.

Bookmarks you made while recording are listed here too. Click one to jump to it, give it a label, or delete it.

Clicking a chapter, a bookmark, a citation in Ask or the longest monologue in Insights also plays the recording from that point, for up to a minute. Click ■ on the playing line to stop it.

Summary

Choose a template with the template button (click it to cycle through them), optionally add Custom instructions such as “focus on the budget discussion”, and click Generate summary. The summary streams in as it is written, and you can Regenerate it at any time.

Summaries always include a title and overview, plus key decisions, action items and open questions when there are any, and your notes. Parley also uses the summary’s title to rename the meeting, unless you named it yourself or it came from your calendar.

Edit summary opens an editor so you can correct or add to it; click Save when done.

Summaries are written by the AI model chosen in Settings → Summary, and are saved as summary.md in the meeting’s recordings folder too.

Templates

Parley comes with two templates:

• Standard Meeting Notes: summary, key decisions and discussion highlights.
• Daily Standup: attendees, yesterday, today, blockers and notes.

You can add your own. Create a JSON file in ~/.local/share/io.github.hankanman.Parley/templates/, such as retro.json:

{
  "name": "Retrospective",
  "description": "What went well, what didn't, and what we'll change.",
  "sections": [
    { "title": "What went well", "instruction": "Things the team was happy with.", "format": "list" },
    { "title": "What didn't", "instruction": "Problems and frustrations raised.", "format": "list" },
    { "title": "Changes", "instruction": "Concrete changes the team agreed to try.", "format": "list" }
  ]
}

Each section’s format is paragraph, list or string. Restart Parley to pick up new templates. The repository has more examples, including project sync, retrospective and client call templates.

Notes

Under the summary, add your own Notes to a meeting. They’re included the next time you generate the summary.

Follow-up email

Draft follow-up email (in the Summary tab) writes a follow-up to send to the attendees, based on the summary and action items. Choose a Tone (Concise, Formal or Friendly), edit the Subject and Body, and copy them into your email app. Parley never sends email itself.

Ask

The Ask tab answers questions about this meeting, such as “What did Priya say about the launch date?”. Answers cite the transcript with timestamps you can click to jump there and hear it. To ask across all your meetings, use Ask Parley.

Insights

The Insights tab shows talk-time statistics worked out from the transcript (no AI involved): the meeting’s length and silence, how much you spoke compared with everyone else, questions asked, interruptions, and the longest monologue. Below that, each speaker’s share of talk time, number of turns and speaking pace.

Calendar event

The header shows the calendar event a meeting is linked to, with its time, location and attendees. Click Link calendar event to pick one from the week around the recording, or Change or unlink it.

Exporting

Export meeting… saves the meeting as a file:

• Markdown, JSON, Word (DOCX) or PDF: summary, action items, notes, attendees and the full transcript.
• SRT or WebVTT subtitles: the transcript only, timed to the audio.

Export to a notes folder (Obsidian)

To keep a copy of every meeting in a notes app such as Obsidian, open Settings → Integrations → Export folder and Choose… a folder, such as your vault. After every summary (and every time you save an edited summary) Parley writes a Markdown note named YYYY-MM-DD Title.md, with the title, date, duration, attendees and tags in its front matter. Sync all now writes or refreshes the note for every meeting that already has a summary, and the line beside it tells you if a note couldn’t be written. Parley only ever updates its own notes, but changes you make to them in another app are overwritten the next time that meeting’s summary changes.

Speakers

Parley works out who is speaking (“diarization”) on your computer and labels the transcript Speaker 1, Speaker 2 and so on. Name a voice once and Parley recognises that person in later meetings.

Speaker detection needs a small speaker model, which Parley downloads automatically. After a recording stops, Parley takes a second look at the whole meeting to tidy up the labels.

Naming a speaker

Click a speaker’s name in a transcript.

1. Choose whether to rename all of that speaker’s lines or only this line (useful when one label really covers two people).
2. Type their name, and optionally their email. If the meeting is linked to a calendar event, its attendees are suggested.
3. Save.

To give the lines to someone Parley already knows, use the target button to pick their profile instead of Create new speaker…. This merges the two, and relabels every line in the meeting.

To undo a wrong name, unassign the lines: they go back to an unnamed speaker.

Note

Parley learns a new voice when you name a speaker in a meeting you’ve just recorded (or imported or re-transcribed in the same session). Naming speakers in older meetings renames their lines but doesn’t save a voice. Naming a single line never changes a saved voice.

Your own voice

On the Speakers page, under Your voice, click Record my voice and read the text aloud for about 20 seconds. Parley then labels you by name in every meeting, including on system audio.

Managing speakers

The Speakers page lists every saved voice. For each one you can:

• Rename it.
• Relearn their voice from all the lines attributed to them, if recognition has been unreliable.
• Merge into… another speaker, if the same person was saved twice. Parley flags profiles with the same name.
• Delete it. Past transcripts keep the name, but future meetings stop recognising the voice.

Re-detecting speakers

Meetings recorded with older versions of Parley can have far too many speakers: the same person split into Speaker 3, Speaker 7 and Speaker 12. To fix one meeting, click Re-detect speakers (the people icon) in its header. To fix them all, click Re-detect speakers in all meetings on the Speakers page.

Parley runs speaker detection again from the saved audio, keeping the transcript text as it is. Any line you’ve named, or that belongs to a saved voice, keeps its name; only the Speaker N lines are regrouped. It takes under a minute per meeting, and Parley tells you how many speakers there were before and after.

This needs the meeting’s audio, so it isn’t available for meetings recorded with Auto-save recordings off.

How recognition works

Voices are matched against saved profiles automatically. When a meeting is linked to a calendar event, people on the invite are matched more readily than people who aren’t, and the number of attendees helps Parley decide how many speakers there are. Voice profiles are stored only in Parley’s local database.

Action items

Parley picks out action items, with owners and due dates when they’re mentioned, whenever it summarises a meeting.

In a meeting

Under the summary, the Action Items card lists the meeting’s items. You can:

• tick an item off when it’s done,
• Edit its text or Delete it,
• Add item by hand,
• Extract from summary (or Re-extract) to pick them out again.

Regenerating a summary replaces the items Parley found before, but keeps any you ticked off and never touches items you added yourself.

All action items

Action items in the sidebar collects items from every meeting, grouped by meeting with the newest first. Filter by Open, Done or All, and click a meeting’s name to open it.

While recording (beta)

With Settings → Recording → Live action items turned on, Parley shows likely action items on the Record page during the meeting, updated every minute or so. These are a preview: the final list is created when the meeting is summarised.

Ask Parley

Ask Parley answers plain-language questions across all your meetings, such as “What did we decide about pricing?” or “Which action items are still unresolved?”. Every answer cites the meetings and moments it came from: click a citation to open the meeting at that point.

To ask about a single meeting, use the Ask tab on the meeting instead.

Narrowing the search

Use the filters above the question box to limit which meetings are searched:

• Person: meetings with a particular speaker.
• Date: the past week, month, three months or year, this year, or a custom range.
• Series: one recurring calendar meeting.

Click New chat to start again. Conversations here aren’t saved once you quit Parley.

Keyword and semantic search

Parley first finds the parts of your meetings that are relevant to the question, then has your AI model answer from them.

• Keyword search always works.
• Semantic search also finds passages that use different words from your question (“budget” for “how much will it cost”). It runs on your computer with a small built-in model (about 146 MB), which Parley downloads in the background shortly after you first start it. There’s nothing to set up.

While Ask Parley is open, Parley indexes new meetings for semantic search in the background. The line under the Ask Parley heading tells you which kind of search is in use and how far indexing has got.

If you use Ollama as your summary provider, Parley uses Ollama for semantic search instead; run ollama pull nomic-embed-text once for it.

To turn semantic search off, or (with Ollama) use a different embedding model, go to Settings → Summary → Ask Parley search.

Answers are written by the AI model chosen in Settings → Summary. With a cloud provider, the matching transcript passages are sent to it.

Calendar and briefs

Connect your calendar and Parley will name meetings after their events, know who attended, remind you before meetings start and prepare a brief from past meetings with the same people.

Adding a calendar

Parley reads calendars as iCal (ICS) feeds, so it works with Google Calendar, Outlook, Fastmail, Nextcloud and most others. To get the feed address:

• Google Calendar: Settings → (your calendar) → Integrate calendar → Secret address in iCal format.
• Outlook / Microsoft 365: Settings → Calendar → Shared calendars → Publish a calendar, then copy the ICS link.

Then in Parley, open Settings → Calendar, paste the URL (https:// and webcal:// both work), optionally give it a Label, and click Add.

Each feed shows when it was last synced. Feeds are fetched when you add them; click the refresh icon to fetch the latest events. Parley keeps events from a week ago to three months ahead.

Note

A feed’s secret address gives read access to your calendar, so treat it like a password. Parley stores it locally and only uses it to download the feed.

What the calendar does

• Upcoming meetings. The next few events in the coming 12 hours are listed in the sidebar under Upcoming.
• Meeting titles and attendees. When you start a recording during an event (from 5 minutes before it starts to 15 minutes after it ends), the meeting is named after the event and linked to it. All-day, cancelled, declined, free-time and solo events are ignored.
• Reminders. A notification 15 minutes (and 5 minutes) before each event, with an Open brief button. Turn these off with Settings → Notifications → Events → Meeting reminders.
• Speaker names. Attendees help Parley name speakers, and can be added to the custom vocabulary.

Briefs

Click an event under Upcoming, or Open brief on a reminder, to see a brief for that meeting. Parley looks through past meetings in the same recurring series or with the same people, and summarises what was discussed, what was decided and what is still open, with links back to the meetings it used.

Briefs are written by your language model. Click Regenerate to write it again.

Importing audio

You can turn any existing recording, such as a voice memo or a call recorded elsewhere, into a Parley meeting with a transcript, speakers and (optionally) a summary.

Importing a file

Click Import audio in the sidebar, or drag an audio file onto the Parley window.

Supported formats: MP4, M4A, WAV, MP3, FLAC, OGG, AAC, MKV, WebM and WMA.

After you pick a file, Parley shows its length, size and format. Set a Meeting Title (it defaults to the file name), then click Import.

Under Advanced options you can set:

• Language: Auto Detect, or one of English, Chinese, German, Spanish, Russian, Korean, French, Japanese, Portuguese, Italian, Dutch, Arabic or Hindi. Choosing the language helps with short or noisy recordings.
• Number of speakers: Auto-detect, or 1 to 8 if you know how many people are talking.
• Model: any Whisper model you have downloaded.

A progress bar shows how far along the import is, and you can cancel at any time. When it’s done, the new meeting opens.

Imports are labelled by speaker too; see Speakers.

Re-transcribing a meeting

If a transcript isn’t good enough, you can run the meeting’s audio through Whisper again, for example with a larger model or with the language set explicitly.

1. Open the meeting.
2. Click the Retranscribe audio button (the circular arrow) in the header.
3. Optionally type a Language code (such as en or de; leave it as auto to detect it), and click Model to cycle through your downloaded models.
4. Click Start.

This only works for meetings whose audio was saved; see Where recordings are saved.

Transcription and AI models

Parley uses two kinds of model:

• a transcription model (Whisper), which turns speech into text. It always runs on your computer.
• a language model, which writes summaries, chapters, action items, follow-up emails and briefs, and answers your questions in Ask. You choose where this runs.

Transcription (Whisper)

Settings → Transcription lists every Whisper model under Model catalog, with its accuracy, size and status. Click Download to fetch one (from Hugging Face), Cancel to stop a download and Delete to remove it. Pick the model used for live transcription under Active model → Whisper model; only downloaded models are listed.

The q5 variants are compressed versions that are much smaller and nearly as accurate. The default, large-v3-turbo-q5_0, is the best balance for most people. Smaller models are faster on slow CPUs; the larger large-v3 models are the most accurate.

Tip

Download large-v3-q5_0 as well as your live model. After each recording, Parley then automatically re-transcribes the meeting with it in the background for a more accurate final transcript. It never downloads a model for this on its own.

Every model understands many languages. To fix the language instead of detecting it, set Settings → Summary → Language & summaries → Transcription language to a language code such as en, fr or de. Leave it blank to detect the language automatically.

Transcription uses your GPU when you installed the Vulkan build (or built Parley with CUDA), and the CPU otherwise. There is no setting for this.

Custom vocabulary

Whisper may misspell names, products and acronyms it hasn’t heard of. Under Settings → Transcription → Custom vocabulary, add each Term the way you want it spelled, optionally with what it Sounds like (for example Kubernetes, sounds like cooper netties).

• The Whisper prompt preview shows exactly what Whisper is given. There is room for roughly a dozen or two terms; Parley warns you about terms that don’t fit.
• Include speaker and attendee names also adds the names of your speakers and of the people invited to the meeting’s calendar event.
• Correct near-miss spellings fixes close misspellings of your terms after transcription.

Custom vocabulary is applied when a meeting is refined after recording, re-transcribed or imported. It isn’t applied to the live transcript.

Language model for summaries and Ask

Choose a Provider under Settings → Summary:

Built-in AI

Runs entirely on your computer with llama.cpp. There are two models:

• Gemma 3 1B (Fast): about 1 GB to download and 1 GB of memory. The default.
• Gemma 3 4B (Balanced): about 2.3 GB to download and 3.5 GB of memory. Noticeably better summaries if your computer can spare the memory.

Download a model and click Select to use it.

Ollama

Install Ollama, then choose Ollama as the provider and pick one of your installed models. You can also type a model name (such as gemma3:4b) and click Pull model to download it from inside Parley.

With Ollama as the provider, Parley also uses it for semantic search instead of its built-in model. Pull the embedding model once:

ollama pull nomic-embed-text

Cloud providers

Paste your key under API keys, then choose a model from the list Parley fetches from the provider. Keys are saved in Parley’s local database in plain text, so anyone with access to your user account could read them.

Warning

With a cloud provider, the meeting transcript is sent to that provider whenever you generate a summary, ask a question or use any other AI feature. Their privacy policy then applies. The built-in model and a local Ollama keep everything on your computer.

Other summary settings

• Templates → Default summary template: the template pre-selected when you generate a summary. See Meeting notes.
• Auto-summarize after recording: write the summary automatically when a recording stops. Off by default.
• Auto-generate chapters after recording: split each new meeting into chapters automatically. Until you change it, it follows Auto-summarize after recording.
• Show transcription confidence: show a coloured dot by each transcript line for how sure Whisper was of it. On by default.

Settings

Open Settings from the sidebar or the tray menu. Changes are saved as soon as you make them.

Notifications

Under General:

• Recording notifications turns the four recording notifications below on or off together.
• Notification sound asks your desktop to play its message sound with each notification.
• Manual Do Not Disturb silences all of Parley’s notifications. Your desktop’s own Do Not Disturb mode also holds them back.

Under Events, choose which notifications Parley shows:

Send test notification checks that notifications reach your desktop.

Call-detection prompts are controlled from Settings → Recording → Meeting detection instead.

Using Parley with Claude (MCP)

parley-mcp is a small Model Context Protocol server that lets AI assistants such as Claude Desktop and Claude Code read your Parley meetings. You can then ask things like “what did we decide about the Q3 budget?” or “which action items do I still have open?”.

It reads Parley’s database directly on your computer, works whether or not Parley is running (even during a recording), and sends your data nowhere except to the assistant that started it. Bear in mind that whatever the assistant reads is then sent to its own model provider.

Getting parley-mcp

parley-mcp comes with Parley (version 0.5.1 and later):

• If you used the install command, it’s already installed at ~/.local/bin/parley-mcp.
• If you downloaded the AppImage yourself, open Settings → Integrations and click Install parley-mcp, which copies it to ~/.local/bin/parley-mcp. After updating Parley, the same button reads Update parley-mcp.

Assistants can’t use the copy inside the AppImage directly, because its location changes every time Parley starts.

Settings → Integrations also gives you ready-made setup commands with the right paths filled in, each with a copy button.

Connecting an assistant

Assistants start parley-mcp themselves, so give them its full path (they don’t understand ~).

Claude Code:

claude mcp add parley -- /home/you/.local/bin/parley-mcp

Claude Desktop: add this to ~/.config/Claude/claude_desktop_config.json and restart Claude Desktop:

{
  "mcpServers": {
    "parley": {
      "command": "/home/you/.local/bin/parley-mcp"
    }
  }
}

Run Parley at least once before connecting, so that its database exists.

What the assistant can do

It can read:

• your meetings, with their summaries, transcripts, chapters, bookmarks, notes and talk-time statistics,
• the calendar event each meeting belongs to, with its attendees, and your upcoming events,
• the people Parley recognises, so it can find every meeting with someone,
• the briefs and follow-up email drafts you’ve made in Parley,
• your action items.

It can search transcripts by keyword, or by meaning, using the same on-device semantic search as Ask Parley. Searching by meaning covers the meetings Parley has indexed, which it does while Ask Parley is open, and needs Parley to have been started once from its AppImage so parley-mcp can borrow its built-in model runner.

It can also add notes, add, edit, complete and delete action items, and replace a meeting’s summary. Anything it creates is marked as coming from the agent.

Try asking it to “prepare me for my next meeting”, “what has Priya committed to this month?” or “draft a recap of every Sprint Planning this quarter”.

The full list of tools is in the parley-mcp README.

Your data and privacy

Everything Parley records and produces stays on your computer. There is no account, no telemetry and no crash reporting. The full details are in the privacy policy.

When Parley uses the network

• Downloading models, from Hugging Face or GitHub. No meeting data is sent.
• Fetching your calendar feeds, if you’ve added any.
• Checking for updates, only when you click Check for updates in the tray menu.
• Listing a provider’s models, when Settings → Summary shows a cloud provider or Ollama.
• Cloud AI providers, only if you chose one in Settings → Summary. The transcript text needed for each request is sent to that provider. With the built-in model or a local Ollama, nothing leaves your computer.

Where everything is stored

If you used Meetily before, Parley moved its data to these locations automatically on first launch.

Backing up

Quit Parley, then copy ~/.local/share/io.github.hankanman.Parley/ (you can skip models/, which can be downloaded again) and ~/Documents/parley-recordings/. To restore, put them back in the same places before starting Parley.

Deleting your data

Deleting a meeting in Parley removes its transcript, summary and related data from the database. Its audio stays in the recordings folder until you delete it there.

To remove everything, uninstall Parley and delete the folders in the table above.

Troubleshooting

Parley won’t start

From a terminal, run parley (or the AppImage) to see any error message.

• “fusermount: command not found” or similar FUSE errors: install your distribution’s fuse3 package, or run the AppImage with --appimage-extract-and-run.
• “GLIBC_2.39 not found”: your distribution is too old. Parley needs Ubuntu 24.04, Debian 13, Fedora 40 or newer.

No sound is being recorded

• Open Audio on the Record page and check the Mic and System level bars move when there is sound. If not, pick a different device from the lists.
• Parley needs PipeWire. Check with pactl info | grep "Server Name", which should mention PipeWire.
• If the other people on a call aren’t being transcribed, make sure System audio is set to the output you’re actually listening on (for example your headset rather than your speakers).

Transcription is slow or falls behind

• Use a smaller or quantised model (the ones ending in q5_0 or q5_1) under Settings → Transcription.
• If you have a GPU, make sure you installed the Vulkan build: reinstall with PARLEY_VARIANT=vulkan. The log says Vulkan GPU support: enabled when it is in use.

The transcript is inaccurate

• Set the Transcription language under Settings → Summary instead of relying on detection.
• Add names and jargon to your custom vocabulary.
• Re-transcribe with a larger model, or download large-v3-q5_0 so meetings are refined automatically after recording.

Summaries fail or are poor

• Check that a model is selected in Settings → Summary. With Built-in AI, make sure its model has finished downloading.
• For better summaries, try the Gemma 3 4B built-in model, a larger model in Ollama, or a cloud provider.
• With Ollama, check that it’s running (ollama list) and that the Endpoint is right.

Ask says “keyword search only”

• “the semantic search model is downloading”: Parley fetches it (about 146 MB) in the background after starting. Leave Parley running with a network connection; Ask switches over by itself.
• “embeddings unavailable” with Ollama as your provider: check Ollama is running and run ollama pull nomic-embed-text. See Ask Parley.

No notifications

Click Send test notification under Settings → Notifications. If it appears, check the Events toggles, that Recording notifications is on (for recording events) and that Manual Do Not Disturb is off. If it doesn’t, check your desktop’s own notification settings and Do Not Disturb mode.

Logs and crash reports

Parley writes a log to ~/.local/state/parley/logs/parley.log (the previous run is kept as parley.log.1). For more detail, start it with:

RUST_LOG=debug parley

If Parley crashes, the log records it, and coredumpctl info parley shows the crash details on most distributions.

Reporting a problem

Open an issue with your distribution, the Parley version (Settings → About), what happened, and the relevant part of the log. Check the log for anything private, such as meeting titles, before sharing it.