Primary workflow
Ask about selected text
Highlight a line in an article and ask a short question. Corivo answers through the lens of your own product and positioning.
← All features
Feature
A private AI assistant whose data path you can actually inspect.
Most AI tools require you to trust their cloud. A private AI assistant inverts that: capture stays on your computer, credentials stay in the operating system keychain, and you can see exactly what the agent sees before it sends anything to a model provider.
The problem
Why a private AI assistant
AI that needs your full inbox, calendar, and codebase to be useful is a security review problem most teams will not pass. Without privacy controls you can inspect, useful AI is a non-starter for regulated work.
Cloud AI tools take your data to be useful. The trade-off is opaque: what is stored, for how long, who can see it. Few users actually read the policy, and the policy can change.
Generic chatbots offer 'memory off' as a switch but no granular controls. You cannot exclude one app, erase one week, or pause for one hour with confidence that something is not being retained somewhere upstream.
Enterprise teams need defensible answers. A private AI assistant with inspectable storage and explicit data flow is the answer for legal, security, and compliance teams that need to certify what data leaves the machine.
What it does
The useful parts, not another chat surface.
No screenshots
Corivo reads structured accessibility text. It does not store screen images, video, or OCR output.
Local storage by default
Capture history and connector tokens live on your computer, with retention and deletion controls.
Pause and exclude apps
You can pause capture and exclude apps that should stay outside Corivo's context stream.
How it works
Privacy as a product constraint
Corivo limits what it reads, keeps local data local by default, and links privacy claims to docs and public desktop source.
- 01Accessibility APIs expose structured text rather than pixels.
- 02Per-app exclusions and pause controls define what enters capture history.
- 03The desktop source is public for users who want to inspect implementation details.
Capture history lives in a local SQLite database on your computer. Inspectable, exportable, deletable. There is no server-side mirror of your capture; the local store is the only store.
Tool credentials are stored in the operating system keychain — macOS Keychain on Mac, Credential Manager on Windows. They do not sit on Corivo servers, and the connector layer never echoes them anywhere outside the keychain.
Data flow when you ask the agent something: query → local retrieval (capture history + tool context) → model provider prompt → response. The Corivo server is not in the retrieval path. The private AI assistant is private at the data level, not just in marketing.
Open source desktop runtime under Apache-2.0. The local-first implementation is inspectable by anyone. The cloud portion handles billing, sync, and team features only — it does not touch the local capture path.
Typical workflows
Where this shows up in real work.
Auditing what Corivo has captured
Open the work memory inspector. Browse capture grouped by app and time. Filter by source, export to JSON, or delete a range. A private AI assistant should make its capture visible; this is how Corivo does it.
Pausing during sensitive work
Working in a financial app or a compliance review? Toggle pause in the menu bar (Mac) or system tray (Windows). Capture stops globally until you resume, with no background re-enable.
Excluding an app permanently
Add an app to the exclusion list. It is never captured, no matter how long you spend in it. Defaults cover password managers and banking apps; you can extend the list to any app you choose.
Erasing a time range
Wipe yesterday's capture if you spent the day on something sensitive. The local store reflects the erase immediately. There are no server-side copies to wait on or trust.
Reviewing what reaches the model
Each agent response can show the prompt context it received. You can verify the exact snippets that left your computer for that single inference call, making the private AI assistant's data path auditable per request.
Use cases on video
Watch the capability show up in real workflows.
Related workflow
Remember what I forgot
Corivo helps recover the specific thing you agreed to yesterday, even when all you remember is the rough scene.
How it differs
How a private AI assistant differs from privacy mode
'Privacy mode' on a cloud AI usually means the conversation does not train the next model. The data still leaves your machine, sits on a server, and is governed by that vendor's retention policy. That is not the same as private.
A private AI assistant keeps the work data on your computer. The only thing that leaves is the prompt for a single inference call, and you can inspect that prompt before it goes. Different posture, different guarantees.
Open source helps. The local-first parts of Corivo are Apache-2.0, which means the data-handling code is inspectable by anyone. The closed cloud parts (billing, sync) do not touch the local capture path, and you can verify that for yourself.
ChatGPT / Claude with memory off
Conversation still goes to a server. A private AI assistant keeps work capture local; only the specific prompt for an inference call crosses the network.
Enterprise SSO + DLP wrappers
Add an audit layer around a cloud AI. A private AI assistant inverts the model: the data is local by default; the audit is built-in.
Self-hosted LLM stacks
Maximally private but expensive to run and integrate. Corivo is local-first for capture and credentials but lets you choose any inference backend, with the data path explicit.
FAQ
Common questions.
What does 'private AI assistant' mean here?
Corivo is a private AI assistant in the sense that work capture, tool credentials, and history stay on your computer. Only the specific prompt for an inference call crosses the network, and you can see exactly what is in that prompt before it goes.
Where does Corivo store my data?
Capture history is in a local SQLite database on your computer at a documented path. Connector credentials are in the operating system keychain (macOS Keychain on Mac, Credential Manager on Windows). Nothing about captured content sits on a Corivo server.
Can I see what Corivo sends to the model?
Yes. Every agent response can be inspected to show the prompt context it received. You can verify the exact snippets that left your computer for that inference call, which makes the private AI assistant's data path auditable per request.
How do I erase what Corivo has captured?
Open the work memory inspector and erase a time range, an app's capture, or the entire store. The change is reflected in the local SQLite database immediately. There are no server-side copies to wait on or to trust.
Is Corivo open source?
The desktop runtime is open source under Apache-2.0. The local-first implementation — capture, storage, retrieval — is inspectable by anyone. The closed cloud portion handles billing, sync, and team features only.
Does pause really stop everything?
Yes. Pause halts capture globally. No accessibility text, no window metadata, nothing flows into the local store until you resume. The toggle is one click in the menu bar (Mac) or system tray (Windows).
How is privacy different on Mac vs Windows?
The storage primitives differ (macOS Keychain vs Windows Credential Manager; the SQLite path differs) but the privacy guarantees and the controls — pause, erase, exclusion — are identical on both platforms.
Related paths
Keep following the thread.
Privacy policy
The legal privacy baseline.
Privacy and storageThe practical version of where your data lives.
PermissionsWhy Corivo asks for each permission.
Work memoryWhat gets captured and how.
Context awareness privacyDetailed notes on what is read.
View desktop sourceInspect the open-source desktop runtime.