O
Oryne

Sign in with your work account

Private executive intelligence across your communications — your account is managed by your company administrator

Forgot password?

Don't have an account? Ask your company administrator to create one for you.

Memories stored on this device
Processed by local AI models
Nothing sent to third-party AI clouds
O
Oryne

System Access Required

Oryne monitors your communications to surface what matters — everything is analyzed on your device by your local AI, never sent away for processing

Accessibility

Read on-screen content to detect conversations and requests

Screen Recording

Monitor activity across Gmail, Slack, LinkedIn, and more

Microphone

Capture voice notes and verbal commitments in meetings

System Audio

Transcribe calls on Google Meet, Zoom, and Teams

Your Inbox

8 open · 0 resolved

Relationships

Who you talk to, who's going quiet, and where follow-ups are owed

Memory Graph — nodes
Relations
Building your memory graph…
Node type
Semantic
Episodic
Working
Factual

Memories

Everything Oryne stores about you — search, inspect, and delete

Loading memories…

Memory Health & Evaluation

Live audit of retrieval quality, graph connectivity, and hybrid search latency on your own memory store

--
out of 100
-- Auditing...
Evaluating your memory layer
Composite index measuring knowledge density, graph connectivity, live retrieval latency, and hybrid (vector + full-text + graph) ranking quality.
Memory Optimization & Health Insights
Evaluating memory signals across connected channels...

Memory Store Composition

Lifecycle mix, capture cadence, and hygiene signals from your actual store — not audited yet

Lifecycle status
Run Audit Health to profile your store.
Memory types
—
Freshness & activity
—

Live Retrieval Benchmark

Execute your own queries against the live hybrid retrieval pipeline and measure latency, hit rate, and top matches across your actual knowledge base

P95 Latency
-- ms
95th percentile across probes
Median Latency
-- ms
50th percentile across probes
Hit Rate
--
Queries returning ≥1 memory
Avg Hits / Query
--
Mean retrieved memories
Evaluation Query Latency Retrieved Hits Top Match Snippet Fusion Score
Click "Run Live Benchmark" to execute real retrieval probes against your memory store.

Agents

Every autonomous worker that acts for this workspace — who owns it, the autonomy tier it holds, and the scopes it may act within.

Autonomy tiers
ObserveReads only. Never mutates.
AdviseProposes actions for a human to approve.
Act (approval)Acts, but every action needs a human approval.
Act (autonomous)Acts unattended within its declared scopes.

Commitments

Everything you said you'd do, tracked across your communications

--
Rescued Follow-ups
-- hrs
Executive Time Saved
--%
SLA Health Index
--
Value Generated
Overdue 0
Due soon 0
Pending 0
Done 0

Documents

Upload PDFs & files — parsed, chunked and searchable. Ask questions with RAG, export as Markdown or Word.

Drop a file here or click to upload
PDF · DOCX · Markdown · HTML · TXT — scanned PDF pages are OCR'd when Tesseract is available
Select a document to ask questions, search inside it, or export it.

Projects

Turn Statements of Work (SOWs) into reviewed project models, Charters (DOCX), WBS workbooks (XLSX), and Kickoff Decks (PPTX).

Projects & SOWs

0
No Project Selected
Select a project from the left or upload a Statement of Work (SOW) to review scope, milestone dates, and generate enterprise deliverables.

Compose

Turn a topic into a tone-aware post for LinkedIn, X, Facebook, Instagram or Slack — with your own voice.

Generated post appears here.

Meetings

Prep cards before, silent fact-checking during, action digests after every meeting.

No upcoming meetings in the window yet.
Select a meeting to open its prep card.

Settings

Manage your account, permissions and preferences

Account
Your Profile
Communication Profile
Connect integrations to build your profile.
Privacy Center
Loading…
System Permissions
Accessibility…
Read on-screen content to detect conversations and requests
Screen Recording…
Monitor activity across Gmail, Slack, LinkedIn, and more
Microphone…
Capture voice notes and verbal commitments in meetings
System Audio…
Transcribe calls on Google Meet, Zoom, and Teams

Grant triggers the real macOS prompt. macOS only allows revoking access from System Settings — Revoke opens the exact pane for that permission.

AI Action Permissions
Loading…

What Oryne may do on your behalf. Anything that leaves this device confirms first; “Ask once” becomes automatic only after you approve a real action. Every grant, confirmation and revocation is audited.

Integrations
Gmail
Email follow-ups, client threads, investor communications
Slack
Team channels, direct messages, engineering updates
Microsoft Outlook
Work & personal mail via Microsoft Graph (OAuth sign-in)
Google Calendar
Meeting intelligence, agenda prep, post-meeting extraction
WhatsApp
Partner conversations, personal connections, quick decisions
Coming soon
LinkedIn
Network introductions, hiring pipelines, board connections
Coming soon
Google Meet
Meeting transcription, action items, verbal commitments
Coming soon
Zoom
Board meetings, investor calls, executive sessions
Coming soon
On-Device Whisper
Checking…
Transcribe voice notes on this device
Audio never leaves your phone — only the finished text is synced
–
When off or not yet downloaded, voice notes fall back to server transcription as before.
Hardware & System Capabilities
Processor & Cores
Detecting…
– Cores
Memory (RAM)
– GB Total
– GB Available
GPU Acceleration
–
CPU
Live Resource Load
Oryne CPU: –%
RAM Used: – GB
Analyzing your hardware configuration…
AI Engine & Billing
Where should Oryne think?
Auto uses Cloud AI when available, otherwise your local engine. Cloud tokens are metered against your plan — local tokens are free.
A loaded model stays inside its RAM budget — the context window shrinks automatically to fit. Applied on the next Start / Load.
Choose How to Run Local Models
Detecting engines on this machine…
Oryne Core Models Preinstalled and auto-installed · everything runs on this device
Loading core models…
Oryne Managed Models Downloaded from Hugging Face · verified & run entirely inside Oryne — nothing else to install
Loading model catalog…
Connect Local Provider
Choose Provider:
Default Ollama port is 11434. Make sure Ollama is running.
Discovered models will populate automatically after running connection test.
Use local model for intelligence & drafts
Process all drafts, briefings, and intelligence 100% locally on your machine
Test Model Inference
Preferences
Draft tone
Tone of AI-generated responses
Auto-dismiss resolved
Remove items from inbox after approval
Notifications
Desktop alerts for critical items
Signature
This will be added automatically when you approve and send.
Extend Oryne
Plugins and developer webhooks
Loading plugins…
Plugins run on every memory, commitment, and draft event. Try commitment_exporter to forward detections to your stack, or build your own via the Plugin SDK (app/plugins/base.py).
Oryne compiles your sentence into a rule you can review before saving.
Loading automations…
Must be a public https URL — private/loopback destinations are rejected (SSRF guard).
Activity History
Memory Dreaming
Dreaming Jobs
No dreaming jobs yet. Memories are formed as you use your integrations.
No activity yet. Approve or dismiss items to build history.
Usage & Billing
Cloud AI is metered against your plan · Local LLM runs on this device and is never charged
Cloud AI Tokens · metered
–
of – weekly limit
Local LLM Tokens · free
–
Never charged · processed on this device
Tool & API Calls
–
this week
Current Plan
–
Upgrades require card payment · downgrades are free
Plans · international card payment (Stripe)
Only Cloud AI consumption (completions, embeddings, cloud speech-to-text) counts against your weekly token quota. Everything processed by Oryne Direct, Ollama or LM Studio stays on this machine and is unlimited. Switch engines any time in Settings → Local AI → AI Engine.
Company
Workspace
Loading…
— …
Domain ownership — DNS TXT
Publish this TXT record at your domain root (@), then click “Verify domain”.
TXT @ — …
Add a user
@yourcompany.com
Members
MemberRoleStatusActions
Loading…
Company AI keys

Store your organization's cloud provider key so members use AI on the company account. Keys are encrypted at rest and shown only masked. If your org policy forces Local AI, these keys are never used. The weekly limit applies per member only while a company key is active.

ProviderKeyBase URLModelStatusActions
Loading…
Add / update a key
Single sign-on (SSO)

Connect one identity provider — OIDC (Okta, Entra ID, Google Workspace) or SAML 2.0. Employees choose “Sign in with company”; accounts are provisioned just-in-time into your existing roles, tenant isolation and IAM policies. A verified domain is required. Credentials (the OIDC client secret, the SAML signing certificate) are stored server-side and shown only masked.

Protocol
Status
Not configured
Redirect URI — register this in your IdP
—
Highest matching role wins. OWNER is never granted by a claim; unmatched users get the default role.
Provider enabled
Allow SSO sign-in through this identity provider.
Enforce SSO
Block password sign-in for company members — they must use their identity provider. Platform admins are always exempt.
Require SSO for members
Role-scoped (company policy): members and viewers must use SSO, but admins and owners keep password sign-in as a recovery path if the identity provider breaks. Applies only while a provider is enabled, and is independent of the blunt “Enforce SSO” above.
SCIM provisioning

Automate the user + group lifecycle from your identity provider (Okta, Entra ID, Google Workspace). Generate a token, paste it into your IdP’s SCIM connector, and set the endpoint below. New hires are provisioned just-in-time, leavers are deactivated, and pushed groups map to Oryne roles via the SSO group → role map. A verified domain is required. The token is shown once and stored only as a hash.

SCIM endpoint — use this in your IdP
—
NameTokenStatusLast used
No SCIM tokens yet.
AI Data Protection

Real-time policy on every AI call. Each prompt is classified for sensitivity (locally, before it leaves) and mapped through the matrix below. Observe records what would happen without changing routing — prove your matrix, then switch to Enforce. Ships Off: nothing changes until you opt in.

Enforcement mode
Off disables the gate. Observe audits every decision. Enforce applies the matrix (scrub / local-only / block).
Output guardrails
Scan model responses for sensitive leakage before they are returned (non-streaming calls). Enforce mode blocks or redacts; audit otherwise.
Per-sensitivity policy
Restricted
Credentials, SSN, financial/medical/legal identifiers.
Confidential
Private company/person data.
Internal
Routine internal content.
Public
No sensitivity detected.
Audit log
Not verified
TimeActorActionResource
No events recorded yet.
Identity & Access

Fine-grained access control. Policies are AWS-style JSON; an explicit Deny always wins, anything not allowed is denied. Everything here is scoped to your company and audited.

My access
Loading…
People

Select a person to see their role and hand them a policy or a group (groups are how a user gains a role’s permissions).

Loading…
Select a person to view and manage their access.
AI Supply-Side Security
Enterprise add-on · opt-in
Status: …
Watches the AI tools your company already uses: secret leaks in synced content, cloud-vs-local AI usage, and control drift from the audit ledger. Off by default — nothing runs until you enable it.
Company Management

Create company workspaces and provision their first administrator. Self-service signup is disabled platform-wide — every account is created here or by a company admin.

Companies
CompanyDomainStatusCreatedActions
Loading…