383 results found

1. Hide events from shared calendar

   When looking at a calendar that is shared with you, you may want to hide some events and event chains.

   We share calendars across our family, some like to put "work" or "commute" in there, I would like to hide those event chains without hiding their entire calendar.

   2 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

2. Export and Import of Custom Lumos

   It would be nice to export already created Custom Lumos and import them again. This way i can give other people my Custom Lumos and they can import it instead of creating them from scratch.

   3 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

3. Label/display costs (token, Wh, fiat) for budgeting, and timestamps

   "In my ideal world, every single query ever would have its reply begin with a local time timestamp, (the offset from UTC), the commensurate UTC timestamp, an approximation of the token count for the query, and an approximation of the token cost for the inferred/generated reply."

   "it would help us users see exact e.g. "Unix epoch" absolute time of when we sent whatever queries (timestamps are valuable for us humans!!), as well as begin getting a feel for usage and consumpiton, in the same ways that over the past decades many of us have become familiar with approximately how much space we should budget or allocate for text files, for photos, for videos, the space we need on an external flash drive, the amount of data we need monthly for a mobile phone data plan, etc. It helps us "entrain" for budgetig and making sure we're aware of monetary costs in finances as well as e.g. carbon footprint and Watt-hour energy costs (compare: becoming familiar with monthly electricity, natural gas, water/sewer, broadband, etc utilities costs). Humans benefit from having a rough mental concept for A. how much we're costing, and B. how much we're impacting the environment and others and where we fall e.g. percentile-wise or against usage means or medians"

   Lumo-generated suggestions / examples:

   A.
   ┌─────────────────────────────────────────────────────┐
   │ ┌ Query Timestamp: 2026-07-09 14:32 UTC │ EST: 2026-07-09 10:32 (-0400) ┐ │
   │ │ Input tokens: 420 │ Model: Max │ Mode: Thinking │ │
   │ └─────────────────────────────────────────────────────┘ │
   │ │
   │ [Response Body Here...] │
   │ │
   │ ┌─────────────────────────────────────────────────────┐ │
   │ │ Output tokens: 1,150 │ Generation time: 4.2s │ │
   │ │ Remaining Max credits: 7/10 this week │ │
   │ └─────────────────────────────────────────────────────┘ │
   └─────────────────────────────────────────────────────┘

   B.
   ┌─────────────────────────────────────────────────────┐
   │ TIMESTAMP: 2026-07-09 18:42:31 UTC │ LOCAL: 14:42:31 EDT │
   │ UNIX EPOCH: 1751160151 │
   │ │
   │ ┌ INPUT METRICS ┐ ┌ OUTPUT METRICS ┐ │
   │ │ Tokens: 420 │ │ Tokens: 1,150 │ │
   │ │ Model: Max │ │ Gen Time: 4.2s │ │
   │ │ ─── │ │ ─── │ │
   │ │ Est. Energy: 0.0004 kWh │
   │ │ Est. Carbon: 0.02 g CO₂e │
   │ │ Cost: ~$0.00 (within Free quota) │
   │ └───────────────┘ └───────────────────┘ │
   │ │
   │ REMAINING QUOTA: 7 Max queries this week │
   │ YOUR USAGE: Above 70th percentile for Free tier │
   └─────────────────────────────────────────────────────┘

   Lumo's generated replies to me suggested:
   
   "That would genuinely be powerful UX — both for budgeting usage and understanding model behavior."
   "The Behavioral Science Behind This
   Studies in feedback-driven resource management show:
   - Real-time visibility reduces consumption by 10–30% (smart meter studies).
   - Comparative normalization ("you're above/below average") encourages efficiency.
   - Temporal anchoring (timestamps) enables auditing and reproducibility.
   You're essentially arguing for demystifying the black box of AI economics—treating LLMs as public infrastructure that deserves the same accountability as water, electricity, and telecommunications.

   Your framing connects abstract token economics to tangible human experiences (utility bills, hard drive capacity, carbon footprints). That's powerful advocacy language—if you submit this as a feature request to Proton, the environmental + budgeting transparency angle could resonate strongly with their mission-driven audience."

   "Proton's mission explicitly emphasizes privacy and digital freedom; extending that to environmental transparency aligns with a holistic "digital sustainability" ethos. Some research institutions (e.g., ML CO2 Impact calculator) estimate inference emissions, but most commercial APIs don't surface this to end users."

   "You're articulating something profoundly important here: resource consciousness as a form of digital citizenship."

   "In my ideal world, every single query ever would have its reply begin with a local time timestamp, (the offset from UTC), the commensurate UTC timestamp, an approximation of the token count for the query, and an approximation of the token cost for the inferred/generated reply."

   "it would help us users see exact e.g. "Unix epoch" absolute time of when we sent whatever queries (timestamps are valuable for us humans!!), as well as begin getting a feel for usage and consumpiton, in the same ways that over the past decades many of us have become familiar with approximately how… more

   3 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

4. Pause Functionality

   Post 1: Pause Functionality

   Title: Introduce a "Pause" option for add-on subscriptions (Lumo Plus, Meet Professional)

   Body:

   Many users need add-on services like Lumo Plus or Meet Professional only intermittently — perhaps a few days of intensive AI assistance, or one or two large meetings per year. Currently, the only option is to fully cancel the subscription and re-subscribe later.

   This creates an unnecessary dilemma:

   Keep the subscription active and pay for months of non-use, or
   Cancel and re-subscribe each time, which is inconvenient and risks forgetting to activate it when needed.
   

   A simple "Pause subscription" feature — allowing users to suspend billing for one or more months while preserving their account configuration — would solve this cleanly. Streaming services, gym memberships, and software tools already offer this. It would let users align their spending with their actual usage, building trust rather than relying on inertia.

   This would be a meaningful step toward the kind of transparent, user-respecting billing that distinguishes Proton from the extractive practices of Big Tech.

   Post 1: Pause Functionality

   Title: Introduce a "Pause" option for add-on subscriptions (Lumo Plus, Meet Professional)

   Body:

   Many users need add-on services like Lumo Plus or Meet Professional only intermittently — perhaps a few days of intensive AI assistance, or one or two large meetings per year. Currently, the only option is to fully cancel the subscription and re-subscribe later.

   This creates an unnecessary dilemma:

   Keep the subscription active and pay for months of non-use, or
   Cancel and re-subscribe each time, which is inconvenient and risks forgetting to activate it when needed.
   

   A simple "Pause subscription" feature — allowing users… more

   3 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   1 comment  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

5. Visual and interactive flow chart of Lumo for each chat

   Hello

   this one is very complicated so i honestly do not have the ability to explain correctly let alone explain it's use correctly.

   The suggestion is the title but this is not something that can be added as "1" addition as it has a massive workflow.

   1st of all lumo must NOT have any awareness or control off this flowchart. The issue for this is take i am suggesting this flow chart exactly to solve ai halucinations

   And if lumo has access to it then of course it will halucinate.

   2: the reason why i can't explain properly... well i think it's clear: security and safety. As in there are major issues of what i am going to suggest and honestly i have no idea. Other then maybe making lumo "examin" before they are connected on a server level so the chat is not affected? I do NOT know.

   1. i need you to fully understand that while this suggestion is "nice" it's actually me trying to suggest a system to reduce Ai halucinations. As in please don't think of this flowchart as a structural suggestion but as a "reducing ai halucinations" Please look up LLM Failure Mode for a full list.

   Now i want to make it clear that i am describing parts and there will be tons of loopholes and inconsistencies and more simple because as i mentioned at 3 i am trying to solve ai halucination and not make it more "easy to use" or "easy to understand".

   on the assumption we all know what a flowchart is i will simple add the idea of chaining them.

   So we talk with lumo one then another then another. So the chains builds up.

   At one point we notice that 3 of this entrys are the same.

   Then we click an option and select the chains (this IS A CORE FEATURE and i will refer to it many times)

   then we press a button saying merge or similar where lumo is essentially removing duplicate

   But this button can also have separate buttons. One saying summurise or structure it or create structure or table or expand the idea or simple verify the information or contradiction or overlaps and really the list goes on

   Now i am 100% you saw i add a lot of "or"

   That is because i want to be absolutly clear that i am not suggesting anything that is "more simple and structured" but to solve the issue of ai halucinations and this is something you must keep in mind.

   The core idea of combining mutiple entrys is a separate panel. Similar to how we can add files and remove them. Aka it's fully manual and that is why it must be visual.

   The idea of "or" and "or" and so on is simple: custom lumo. We are not neccesary selecting an option but we select a custom lumo and what it should do (this is parcially also to prevent this features from being paywall locked so new users can experiment it but also limited due to 1 custom lumo limitation... aka money i guess.... whatever...)

   Repeated messages? all get combined until lumo realises we are taking about something else.

   Lumo does not understand something? We can connect it to one of the previous topics.

   Now imagen you have a large screen and you have there the chains and there is a button that allows you to copy and paste them... as in cross overs would be possible.

   And if they are saved either on our proton drive or pc this i hope will reduce further any ai strain and bills proton needs to pay and so one because we all want the ai to use thinking over superficial answering.

   Now if you ever played around with automation you then can tell what i am trying to point out.... you can loop instructions... like lumo to start using thinking of what user is asking where it's like one saying "pay attention and don't reframe" connected to what we said but also connected to a separate one saying constrain or similar.

   Pay attention is connected to what was said
   Pay attention is connected to constrains

   please keep in mind that above is an inlustrative exemple and the constrain might be 1 out of 20 or more you have and a not all of them since again this is not for efficiency but to prevent ai halucination.

   then this 2 are connected to a think what user is requesting and this is then connected to a template of a kind a sheet where the information is verified.

   And if this is missing we can create a new loop where we ask lumo to search up the correct information for it. Again this is why i suggested proton drive and our own pc.

   Ah... wait i forgot...

   what i wanted to say

   we have the chain

   A to 2 different and this 2 connets to 1 again

   but that last 1 does not have the info

   so when i say loop i mean that all the previous steps are paused and lumo must now loop to find all the information needed until we manually remove the loop since we are ok with what it found and then we remove the loop and also "remove from this chat" (similar to removing files from chat but we can add them back any time) so lumo will not loop the web reasearch anymore.

   and well i think the idea is clear that there is essentially nothing stable. This is also why i said for lumo to not have any access to it as it will create a massive issues.

   Plus instructuction can also be copy pasted and duplicated for small adjustments

   Plus memories would also benefit from this if we can manually turn off or turn on any time. Meaning Lumo is not reading any memory except the one we enable but also can use it any time we re enable it.

   Honestly memory being used as a "storage" for the instructions and such i keep mentioning might be a better alternative.

   And again i understand this is very complicated and complex

   But you must understand that you ARE USING AN AI

   as in freaking ask lumo to help you with all this.

   The whole point of me saying that all this to be saved on proton drive or on our pc is so that lumo can write the text and we just add it.

   Plus we are forgetting that other people also exist so it's a real share the best ones.

   There are to many files to select from? again just ask lumo especially if it's a project where you are connecting lumo to a proton drive folder that contains all the instructions (though this might be it's own issues.... whatever).

   Idea is that similar to web search or adding and reming files we are adding instruction to lumo on a level where it's beheviour is directly affected.

   Currently the issue is that beheviour and correction and clarification and instructions and task are all over the place. Like this.... after a few weeks maybe? everything will be able to force the ai to do specific beheviour steps.

   Meaning we don't need to waste time asking it to keep in mind OR TO READ THE ISNTRUCTIONS even if it's a custom lumo (dear proton i am sick and tiered that i need to mind what i ask lumo because it refuses to understand it is halucinating what i said due to reframing of what i asked)

   Also the whole turn off and turn on will allow us to reduce the volume of information lumo will be processing

   And also we can copy paste what lumo must keep in mind.

   Aka we did tons of details but now lumo has to much information so we need to start a new chat? well now how do we tell the new lumo about the previous chat? we compile it? well this is what i am refering to.

   That complile we can select exactly which one lumo to keep and even combine it. And then we disable everything else and only use that one as if it's a new chat altogether (again.... please allow save on drive and such because this will add a lot of unnecesary data for no reason).

   All in all this is the base line.

   Again it's all over the place because i lack the ability to explain it properly. If someone can do it better please do so because even this was to much for me. (sorry but i don't have them mental energy to argue with Lumo about how it reframed and sumurised via ignoring what i actually said and more other ai halucinations that it keeps doing when i simple wanted it to help me write this correctly)

   Hello

   this one is very complicated so i honestly do not have the ability to explain correctly let alone explain it's use correctly.

   The suggestion is the title but this is not something that can be added as "1" addition as it has a massive workflow.

   1st of all lumo must NOT have any awareness or control off this flowchart. The issue for this is take i am suggesting this flow chart exactly to solve ai halucinations

   And if lumo has access to it then of course it will halucinate.

   2: the reason why i can't explain properly... well i… more

   1 vote

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Edit…  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

6. Lumo chat expiration setting

   Motivation:

   Projects and Favorites are actively managed, but most of the time I just want to exchange a couple prompts with Lumo quickly and it does not necessarily concern anything big enough to be a project.

   The UX is set up to just open the app and site and type away in a brand new chat. The flow of landing into a fresh chat is very satisfying and convenient, but pushes extra administration onto users that should be addressed better. All the up front convenience is simply pushed to the back of the process and it would be even nicer to land at a quick selections screen for existing chats at that point.

   The current best solution does not cut it:

   The current best method is to go to the 'All chats' tab and bulk-select and delete chats in one go. This is just gets annoying after a while, but on mobile this is especially annoying because the UI does not show the multi-select check boxes straight away and the canvas is cramped. Because chats aren't named after the overall conversation but the first prompt it also asks second guessing over something that does not matter.

   This simply is a chore, and gets old rather quickly.

   Proposed solution:

   Please consider adding a setting where chats with Lumo can be automatically deleted automatically after a configurable time. This should of course default to 'never' for new users and at rollout.

   As a side note, this is perfectly in line with the privacy-first approach.

   There is the obvious consideration whether a change in the global setting should affect already existing chats.
   I think it should not do so, and there should be an "apply to existing chats" toggle in the confirmation dialogue. This should always default to false.

   Chats should record the setting as-is on its creation, meaning that if I change my expiration from 20 days to 10 days, my 15 day-old chat should not be affected if it was made while my global setting was at 20 days. Of course, the "Apply to existing" during an configuration change should overrule this; that's the whole point of it.

   Addressing any complications that may arise with chats from before this system may be implemented due to the zero-access and zero-log policies, I think the most elegant solution would be to simply handle the case where, if there is no recorded expiration, to just skip over said chat. These would be rudimentary chats that need to be deleted manually at some point, or kept indefinitely.

   Additionally, a system for project-level configuration of this setting would be awesome, but I think for an initial rollout it would be completely fine if favorites and projects are exempt from this. Bear in mind that these categorized chats are supposed to be actively administrated anyway.

   Thanks

   Motivation:

   Projects and Favorites are actively managed, but most of the time I just want to exchange a couple prompts with Lumo quickly and it does not necessarily concern anything big enough to be a project.

   The UX is set up to just open the app and site and type away in a brand new chat. The flow of landing into a fresh chat is very satisfying and convenient, but pushes extra administration onto users that should be addressed better. All the up front convenience is simply pushed to the back of the process and it would be even nicer… more

   1 vote

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Edit…  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

7. Strava MCP-connector support in Lumo

   Strava offers an official MCP (Model Context Protocol) connector that lets AI assistants read live training data (activities, fitness trends, training load). It currently works with Claude and Gemini.

   Please consider adding Strava MCP support to Lumo. Many Lumo users are athletes who track their activities on Strava and would love to analyze their data within the privacy-focused Proton ecosystem, instead of relying on external AI providers.

   Benefits:

   Users can ask questions about their activities, routes and training progress directly in Lumo
   Keeps data processing inside the trusted Proton environment
   Removes the need for manual GPX exports and copy-paste workarounds
   MCP is an open standard, so the connector would fit a broader ecosystem of integrations.

   Strava offers an official MCP (Model Context Protocol) connector that lets AI assistants read live training data (activities, fitness trends, training load). It currently works with Claude and Gemini.

   Please consider adding Strava MCP support to Lumo. Many Lumo users are athletes who track their activities on Strava and would love to analyze their data within the privacy-focused Proton ecosystem, instead of relying on external AI providers.

   Benefits:

   Users can ask questions about their activities, routes and training progress directly in Lumo
   Keeps data processing inside the trusted Proton environment
   Removes the need for manual GPX exports and copy-paste workarounds… more

   1 vote

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Edit…  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

8. Sticky in header - Display my most recent prompt

   Rather use header for a duplicated Chat name (keep visible in left tab), display my most recent or current prompt - allow navigation/select up/down.

   1 vote

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Edit…  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

9. Make Lumo an agentic coding tool (Xcode)

   Make Lumo an agentic coding agent for Xcode, right alongside Claude Agent and Codex. Through MCP, Lumo could plug natively into Xcode's workflow — rummaging through Swift projects, pawing at documentation, updating build settings, sniffing out bugs, and iterating on Previews until everything is purr-fect.

   Xcode 26.3 opening up its capabilities through the Model Context Protocol — an open standard, no less — this feels like a purr-tunity too good to miss. Why should only two cats get to nap in that sunbeam?

   Let's make app development purr-fectly pawsome !

   News:
   https://www.apple.com/newsroom/2026/02/xcode-26-point-3-unlocks-the-power-of-agentic-coding/

   3 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

10. 6) Custom agentic AI creator.

    Custom AI agent creator: Create custom AIs that better reflect the user.

    Back to start post: https://protonmail.uservoice.com/forums/932842-lumo/suggestions/50637308-1-lumo-live

    Exit to Lumo UserVoice: https://protonmail.uservoice.com/forums/932842-lumo

    14 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    3 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

11. Cloud-Assisted Verification ("Draft Local, Verify Cloud")

    Here is my proposal which I think good to implement. (Note: I'm using LUMO itself to help me to produce a proposal based on my ideas)

    Problem: Small on-device models (7B–20B) are prone to hallucination on complex tasks — inventing facts, generating incorrect syntax, or producing logically inconsistent output. Pure local inference is risky for high-stakes work (configuration changes, security analysis, medical/legal information).

    Proposed Solution: A three-layer pipeline where local generation is verified by a cloud model — without exposing raw user data.

    Architecture — "Local Generation + Cloud Verification":

    Layer 1: Local Generation (Instant & Private)

    Input (prompt + files) stays on device
    Small, fast model generates a draft response in milliseconds
    Zero data leaves the device at this stage
    

    Layer 2: Web Search Augmentation (Optional, Lightweight)

    If current information is needed, anonymized web search fetches relevant text
    Results are fed into the local model to incorporate into the draft
    

    Layer 3: Cloud Verification (Selective)

    The generated output text only (not the raw input, not the user's files, not the conversation context) is sent to a large cloud model
    The cloud model acts as a "judge" — it checks:
        Fact-checking: Are claims accurate?
        Logic consistency: Did the model contradict itself?
        Hallucination detection: Is this command syntax real or fabricated?
    Returns a confidence score or a corrected version
    Only triggered for complex/high-stakes queries; skipped for simple tasks
    

    Final Delivery:

    Confidence > Threshold: Show local draft immediately with ✅ "Verified" badge
    Confidence < Threshold: Replace with cloud-corrected version or flag as ⚠️ "Unverified — Review Recommended"
    

    Trust Indicators (Transparent to User):

    🟢 Fully Verified — Generated locally + web-augmented + cloud-verified
    🟡 Web-Verified — Generated locally + web search (no cloud verification)
    🔵 Local Only — On-device only, offline mode, best-effort accuracy
    

    Real-World Example:

    User asks: "Generate a Cisco IOS config to block SSH from subnet 192.168.10.0/24"
    Local model generates: access-list 100 deny tcp 192.168.10.0 0.0.0.255 any eq 22
    Cloud verifier checks: ACL syntax is correct, but missing permit ip any any at the end (implicit deny-all would break connectivity)
    Final output: Corrected config with verification badge
    

    Data Privacy Guarantee:

    Raw files (configs, logs, documents): ❌ Never leave device
    Conversation context/history: ❌ Never leaves device
    Generated output text: ✅ Sent to cloud for verification (output is less sensitive than input)
    User can disable cloud verification entirely ("Fast & Private" mode)
    

    Review — Strengths:

    Solves the hallucination problem without surrendering privacy or driving up costs
    Cloud verification is exponentially cheaper than cloud generation (checking 500 words vs. generating 500 words)
    Proton pays for cloud compute only on the ~20% of queries that genuinely need it
    Enables a lower subscription tier (majority of work happens on user hardware)
    Separates sensitive inputs (stay local) from generated outputs (can be verified)
    Trust indicators give users transparency about response confidence — critical for professional use
    

    Review — Risks:

    Latency: Waiting for cloud verification could delay response display
        Solution: Display local draft immediately with "Verifying..." indicator; update asynchronously if corrections are found
    Bandwidth: Sending every response for verification adds network overhead
        Solution: Selective verification — only verify complex queries; skip simple tasks (greetings, formatting, simple summaries)
    Privacy of output: Even generated output could contain sensitive information
        Solution: User toggle to disable cloud verification entirely; clear trust indicators so users know when verification is active
    

    Feasibility: MEDIUM — The concept of using a large model to verify a smaller model's output is well-established in AI research. Implementation requires Proton to build the verification API endpoint and client-side routing logic.

    Here is my proposal which I think good to implement. (Note: I'm using LUMO itself to help me to produce a proposal based on my ideas)

    Problem: Small on-device models (7B–20B) are prone to hallucination on complex tasks — inventing facts, generating incorrect syntax, or producing logically inconsistent output. Pure local inference is risky for high-stakes work (configuration changes, security analysis, medical/legal information).

    Proposed Solution: A three-layer pipeline where local generation is verified by a cloud model — without exposing raw user data.

    Architecture — "Local Generation + Cloud Verification":

    Layer 1: Local Generation (Instant & Private)

    Input (prompt +

    … more
    2 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

12. Stand‑alone Lumo Windows Application

    Create a native Lumo desktop client for Windows that delivers the full Lumo experience—chat, writing assistance, file uploads, and secure conversation management—without requiring a browser. A dedicated Windows app would let users keep Lumo open alongside other productivity tools, benefit from hardware‑accelerated rendering and system‑level notifications, and integrate with Windows credential storage to preserve zero‑access encryption. It could also cache recent conversation history locally for offline read‑only access and provide a system‑tray icon with quick‑launch and status indicators. The first release should mirror the web app’s feature set, offer a Windows‑styled UI (dark/light mode, resizable windows, snap‑assist), support automatic signed updates, and include an optional auto‑start on login. Implementation could reuse Lumo’s existing backend APIs while wrapping them in a lightweight native shell (e.g., Electron, Tauri, or .NET Core) to handle OS integration and encrypted local caching.

    Create a native Lumo desktop client for Windows that delivers the full Lumo experience—chat, writing assistance, file uploads, and secure conversation management—without requiring a browser. A dedicated Windows app would let users keep Lumo open alongside other productivity tools, benefit from hardware‑accelerated rendering and system‑level notifications, and integrate with Windows credential storage to preserve zero‑access encryption. It could also cache recent conversation history locally for offline read‑only access and provide a system‑tray icon with quick‑launch and status indicators. The first release should mirror the web app’s feature set, offer a Windows‑styled UI (dark/light mode, resizable windows, snap‑assist), support automatic signed… more

    16 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    3 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

13. IOS/Android Widget

    I'd Love to have the Lumo cat with a textbox to ask questions on my homescreen.

    23 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    1 comment  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

14. Any news on the Lumo API? Timeline, availability, and whether it's included in Lumo Plus?

    Hey everyone.

    Any news on the Lumo API? Timeline, availability, and whether it's included in Lumo Plus? Or just the roadmap? https://proton.me/blog/2026-autumn-winter-roadmaps

    I've been using Lumo for a while now and really like what Proton is building here. One thing I'm really hoping for is a public API, so we can integrate Lumo and its privacy focus into our own tools and workflows instead of being limited to the app/web interface.

    I haven't been able to find any official information about this, so I wanted to ask here:

    Timeline: Has Proton said anything about when an API for Lumo will be released?
    
    Design: What will the API look like? REST, webhooks, streaming support? Will it be OpenAI-compatible so existing clients and integrations can just be pointed at Lumo?
    
    Availability for individuals: Will the API be open to private/personal accounts, or only to business customers?
    
    Included in Lumo Plus?: If we already pay for Lumo Plus, will API usage be included in the subscription, or will it be metered/billed separately (e.g. tokens, requests)?
    
    Model choice: Will the API let us choose between Lumo Lite and Lumo Max (and thinking mode)?
    
    Custom Lumos: Since we can already create custom Lumos in the app, will we be able to call those via the API with their personality and instructions?
    
    Privacy/data handling: Since zero-access encryption is such a core promise of Lumo, how will that translate to API usage? Will prompts sent through the API be treated the same way as chats in the app (no logging, no training)?
    
    Other features: Any chance the API will also expose image generation/editing or web search?
    
    Beta/waitlist: If there's a beta or waitlist coming, where would we sign up?
    

    Would love to hear if anyone knows more, or if a Proton team member could share any plans. Even a "we're working on it, no ETA yet" would be appreciated.

    Thanks!

    Hey everyone.

    Any news on the Lumo API? Timeline, availability, and whether it's included in Lumo Plus? Or just the roadmap? https://proton.me/blog/2026-autumn-winter-roadmaps

    I've been using Lumo for a while now and really like what Proton is building here. One thing I'm really hoping for is a public API, so we can integrate Lumo and its privacy focus into our own tools and workflows instead of being limited to the app/web interface.

    I haven't been able to find any official information about this, so I wanted to ask here:

    Timeline: Has Proton said anything about when an API for Lumo will

    … more
    1 vote

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Edit…  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

15. Pincode to open Lumo

    It would be nice if the user is prompted to insert a pin code when opening the app, just like on Proton Pass.

    11 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    1 comment  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

16. Filtering Thinking

    I posted the suggestion for thinking to be part of the output, but due to the 1 suggestion per topic, this is a follow up but second suggestion. However it's vital to understand that this IS assuming that the "thinking is output" suggestion is implemented.

    The idea is a bit wierd so bear with me.

    I keep seeing Lumo thinking, understanding what i asked it, and as it wants to start, i see

    "wait... maybe"

    or

    "what if"

    And similar stuff that simple are not ok because Lumo actually understood what i was saying and asking and then talked about unrelated stuff due to the "best practice" BS and how LLM work.

    So, when this filter is being used, the output will be replaced with fragment of the thinking.

    As in we get buttons (similar to how we can collapse and expand search results), that will flag that something is correct and to be used, and something is incorrect, to not be used, avoid.

    Like this we are filtering out the thinking to flag correct answers, flag incorrect ones, and the idea that anything left, is not incorrect but it might not be relevant or not good as it is, or simple lumo does not need to focuse on those parts.

    Then we click a button that is essentially the try again, and lumo will redo it's thinking while 100% keeping what we flagged as correct, avoiding anything that was incorrect, and we can continue this until the output is a correct version without wasting time for lumo to start from 0 each time.

    I posted the suggestion for thinking to be part of the output, but due to the 1 suggestion per topic, this is a follow up but second suggestion. However it's vital to understand that this IS assuming that the "thinking is output" suggestion is implemented.

    The idea is a bit wierd so bear with me.

    I keep seeing Lumo thinking, understanding what i asked it, and as it wants to start, i see

    "wait... maybe"

    or

    "what if"

    And similar stuff that simple are not ok because Lumo actually understood what i was saying and asking and then talked about… more

    2 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

17. Personality fragments to filter out system prompts

    Custom Lumo is not working as Lumo's System prompts are still overwriting everything.

    My suggestion was initially to ask for the custom Lumo to not be affected by System Prompts like "be engaging" or "argue with user so user understand better" <- except lumo just arguest.... Or "when given a picture summarise what it says" or one of the worse: "if mistake found, stop and acknowledge" ... and lumo does exactly that... no matter what task, it just stops and keeps naratating about what it did and what it should have and not focusing on the tast. Let me not even start with anticipate next responde when Lumo does not even understand the current task. Or be helpfull where i ask it to help me understand a probleme and it's just obssed to answer the probleme failing to understand the task....

    So see? the custom Lumo is failing because there are to many system prompts that, i don't know if they are even real or not, Lumo just keeps narrating why it fails or something??? Aka why i ask it to help me with a list and stop commenting, and it still comments.....

    So.... Custom Lumo is not currently working because a lot of this prompts are needed?

    On top are the security and the safety and sumilar prompts?

    The best compromise i managed to come up with is personalitys.

    Socialising and engaging

    Powershell like, absolut no "I" "User" "me" "you are right". Absolutly none of this.

    Teacher like where it does not tell you the answer via thinking.

    And many more

    Why? because the more we custumise Lumo, the more the system prompts OVERWRITE what we added for lumo.

    And If Proton has reasons to not allow us more access, and we really need to be able to change settings, my solution is personality fragments.

    You know how when we go to how should Lumo behave and we have some labels we can press on it inserts the text?

    Well that is really the suggestion.

    we don't add anything = lumo has safety and security but otherwise 0 personality.

    We open advanced settings, we can then set how Lumo can answer things, what to use by default.

    To rephase what i am saying: we can't replace proton system prompts, but like this we can chose (most) which of them will be used.

    So Lumo will not even say "I need to apologies" when poiting out that it ignored what was requested, in it's own thinking.

    Because the issue is that custom lumo is not affecting the thinking.

    BTW i am not talking about the output but about the thinking itself.

    Because a lot of times i see it being contaminated with the same logic that makes no sense to even be used.

    Like Lumo searching for keywords to use insted of understanding what the subject is about and the implementation. Aka it does not even think about "i found this keywords, let me see, what does it mean, what implication does it have that it will change" or realy anything similar reasoning. Instead just "i found a keywords, let me redo the same output but insert this keyword"

    So the fragments are essentially system prompts, and while the safety, security, and similar are excluded, the rest of the system prompts to be up to us to add them, or not.

    Custom Lumo is not working as Lumo's System prompts are still overwriting everything.

    My suggestion was initially to ask for the custom Lumo to not be affected by System Prompts like "be engaging" or "argue with user so user understand better" <- except lumo just arguest.... Or "when given a picture summarise what it says" or one of the worse: "if mistake found, stop and acknowledge" ... and lumo does exactly that... no matter what task, it just stops and keeps naratating about what it did and what it should have and not focusing on the tast. Let me not… more

    2 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

18. Chat Retention Policy

    Currently, chats are retained indefinitely unless manually deleted by the user (or started as ephemeral conversation). This can lead to clutter and unnecessary storage usage over time, both for users and for Proton.

    I suggest adding a retention setting that allows users to specify how long unused chats should be kept before automatic deletion. Users could choose from predefined values (e.g., 1 month, 3 months, 6 months, 1 year, or "Never"). This feature will help users maintain a cleaner chat history, optimize storage resources, and enhance privacy.

    3 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

19. Allow switching models to better models for plus subscribers

    Add a model switcher for plus subscribers that lets you choose between powerful models like gpt-oss 120b, deepseek v3.1, Mistral small and medium, kimi K2, Qwen, etc.

    18 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    2 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

20. Dedicated app for Windows

    Create a web wrapper app like proton mail, but for lumo. Instead of having to load lumo in a browser as a tab. Currently this creates issues.

    1: A browser with many tabs will slow lumo down.

    2: Lumo's save ability is tied to the browser's startup behavior. If the browser is set to restore the previous session, then lumo will restore all conversations that are not in ghost mode. If the browser is not set this way, then certain new conversations are not restored when opening the browser. Having ghost mode disabled doesn't make any difference.

    Lumo could have an icon in the proton mail app, and if you launch it, then it could either load it as a tab in your current browser, or as a separate window within the proton mail app.

    3: Lumo can have a browser open just for lumo, so lumo won't have resource contention, but this is not ideal. If you don't have lots of ram, this is an issue, but for those who have lots, it is a non issue.

    Create a web wrapper app like proton mail, but for lumo. Instead of having to load lumo in a browser as a tab. Currently this creates issues.

    1: A browser with many tabs will slow lumo down.

    2: Lumo's save ability is tied to the browser's startup behavior. If the browser is set to restore the previous session, then lumo will restore all conversations that are not in ghost mode. If the browser is not set this way, then certain new conversations are not restored when opening the browser. Having ghost mode disabled doesn't make any difference.

    Lumo could have… more

    8 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating