OUTCRYSERVER
Talk to us→ Request a demo
For data that cannot leave

For teams whose data cannot go to hosted AI.

Run Outcry behind your firewall, connect over LAN or VPN, and keep sensitive work on organization-controlled hardware.

Request a demoHow it works
No cloud inference No seats No metering
Looking for the free individual app?
1 Mac. Whole org.
Run inference on a single appliance, serve every organizer who needs it.
OpenAI-compatible
Existing tools and SDKs work — change the base URL, change the key, ship.
Pinned releases
Admin promotes model updates. Production never silently swaps under you.
No seats. No metering.
Flat org license. Invite as many internal users as the hardware can handle.
The product line

Two products. One architecture.

Both run Outcry locally — no hosted inference. Pick the one that matches your scope.

Outcry · Free

Outcry App Store · free

For individual organizers. Runs the local model on one personal device.

  • One Mac, one user — your laptop or desktop
  • App Store install, sandboxed, auto-updating
  • Local model, private by default, no account required
  • No API keys, no admin chrome, no team controls
Get Outcry on the App Store →
Outcry Server · Paid

Outcry Server Org-wide · founder-led

For organizations. One customer-controlled Mac serves your whole staff over LAN or VPN.

  • One Mac mini or Mac Studio, your hardware, your network
  • OpenAI-compatible /v1 endpoint for staff & internal tools
  • Per-tool API keys, pinned model updates, metadata-only logs
  • Onboarding, support contract, no seats, no metering
Request a demo →
Hosted AI vs. Outcry Server

Where inference runs decides the rest.

If a CISO or executive director has 60 seconds, this is the table that decides fit.

Axis
Hosted AI
Outcry Server
Where inference runs
Vendor datacenter
Your Mac, your network
Who controls logs
Vendor retention & subpoena posture
You — off, hashed, or audit, admin's call
Who controls model updates
Silently swapped behind the API
Pinned — admin promotes when ready
How billing scales
Per seat, per token, per month
Flat org license — no seats, no metering
Network exposure
Outbound to third-party endpoints
LAN or private VPN only
Why local · why now

Your staff needs AI.
Your data cannot leave.

Source identities. Privileged communications. Whistleblower intake. Coalition strategy. The work that hosted AI providers are most hungry for is exactly what you cannot send to a third-party datacenter — and yet your team is already pasting it into ChatGPT.

There is another way.
How it works

Five steps. One Mac. Everyone connected.

From App Store to organization-wide AI in an afternoon. No cloud. No seats. No metering.

01 Download

Free Outcry app

Anyone can install the free Outcry app from the App Store. Individual organizers run the local model on their own device.

APP STORE
02 Install

Outcry Server

A signed, notarized direct download for Mac mini or Mac Studio. First-run wizard handles install, model import, network choice.

03 Pin

Lock the model

Server detects, verifies, and stages new Outcry model releases. Production stays pinned until your admin explicitly promotes.

v3.1 LIVEv3.2 STAGED
04 Connect

Team comes online

Staff and internal tools connect over LAN or VPN using the OpenAI-compatible /v1 endpoint. Remote staff via Tailscale-style VPN.

05 Govern

Admin controls

Issue keys, revoke access, view metadata-only logs, manage updates and diagnostics — all from a simple GUI. No CLI required.

The product

One status. One toggle. Advanced when you need it.

Outcry Server · Dashboard outcry.lan/v1 · 14 active
Private endpoint online
14 active requests
Mac Studio · M2 Max · 64 GB · uptime 18d 4h
https://outcry.lan/v1Copy
Drop-inPoint your existing OpenAI client at this URL — Cursor, Raycast, the OpenAI SDK. Same key, same calls.
Model
outcry-72b
Tok / sec
38.4
Active keys
7
Privacy mode
METADATA

Last 24 hours

Requests12,840
Avg latency412 ms
Errors2

Connected clients

Legal team · Cursor5
Comms · Raycast3
Casework · custom6

API keys

Issue keys to teams or tools. Revoke access without rotating the rest.

NameKeyLast usedStatusActions
Legal teamsk-out-legal-•••••••••••a4f22 min agoActive
Comms · Raycastsk-out-comms-•••••••••a8b138 min agoActive
Casework intakesk-out-cases-••••••••••e72djust nowActive
Staging · Mac minisk-out-stage-••••••••••71fc3 days agoActive
Old contractorsk-out-cntrc-••••••••••91a812 days agoRevoked
Production · Pinned
outcry-72b · v3.1.4
Released Mar 8 · 218,000 prompts served · 0 model swaps in 92 days.
Hashsha256:7f3a…b81c
Size41.8 GB
Promoted bym.bornfree
Staged · Awaiting promotion
outcry-72b · v3.2.0
Detected May 1, verified May 1. Test suite passed. Promote when ready — production is not affected.
Hashsha256:c4d2…9e07
Size43.2 GB
Test suitePASSED · 247 / 247

v3.2 · changelog

  • Improved long-context campaign analysis (context window 128k → 200k tokens)
  • Tighter coalition-mapping outputs; less hedging on tactical recommendations
  • Adds 340 newly-indexed primary-source movement texts (1968–2024)
  • Reduces hallucination on legal-precedent retrieval by 41% on internal eval
What you get

A working AI appliance, not a parts list.

Outcry Server ships as a complete, supported system — installer, GUI, OpenAI-compatible runtime, admin chrome, and a year of updates.

Private local inference

Prompts and responses are processed on the customer-controlled Mac. Never leaves your hardware unless you route it there.

OpenAI-compatible

Use existing tools and SDKs by changing the base URL and API key. /v1/chat/completions, streaming, embeddings.

No seats. No metering.

Flat org license. Invite as many internal users as the hardware can handle. Your finance team will love this.

Simple GUI

A WARP-like menu bar app: one status, one toggle, advanced settings hidden until you need them. No Terminal required.

Keys & logs

Issue keys to teams or tools, revoke without rotation, view usage. Prompt contents are not stored by default.

Pinned model updates

New Outcry releases are detected, verified, and staged. Admin promotes when ready — production is never silently swapped.

Onboarding & support

First year includes server updates, compatibility releases, documentation, hands-on onboarding, and direct support.

Production + backup

One production server license plus one backup/test instance. Stage updates somewhere safe before you ship them.

VPN-ready

Built for LAN by default; tested with Tailscale-style private networking for remote staff. We do not recommend public exposure.

Built for

Organizations whose data cannot go to hosted AI.

If sending prompts to a third-party datacenter is a non-starter — for legal, ethical, or operational reasons — this is for you.

Legal aid

Client-privileged casework

Privileged conversations, intake notes, and brief drafting that cannot enter a vendor pipeline.

Privilege · Intake · Drafting
Newsrooms

Source-protected reporting

Source identities, draft scripts, and unpublished investigations that never leave the newsroom network.

Sources · Drafts · Search
Unions

Coalition strategy

Bargaining strategy, member data, organizing models — kept inside the union's own infrastructure.

Strategy · Members · Comms
Advocacy

Campaign planning

Power maps, escalation timelines, and donor data that should not feed a third-party model.

Mapping · Comms · Research
Sanctuary

Whistleblower workflows

Whistleblower intake, immigration support, and shelter coordination — operationally safe to keep on-prem.

Intake · Support · Coord
Good fit

You'll get value from this.

  • Org-controlled hardware (or budget for a Mac mini / Mac Studio)
  • Regulated, privileged, source-protected, or otherwise confidential data
  • 10–500 internal users on LAN or private VPN
  • Existing OpenAI-compatible tooling — Cursor, Raycast, internal apps
Most of our customers are legal aid orgs, newsrooms, unions, and advocacy groups in this shape.
Probably not yet

We'll save you the call.

  • You need public, internet-facing inference (consumer apps, embeds)
  • No Mac hardware budget and no plan to acquire one
  • Single-user or hobby use case — one organizer, one device
  • Looking for usage-based billing or ChatGPT-style seats
If you're a single organizer, the free Outcry app is the right tool — same model, runs locally on your own device.
Pricing

Flat org pricing. No seats. No metering.

Movement-grade fairness: one price, every organizer, no per-message arithmetic.

First-year package
Outcry Server
From $6,000/ first year

One flat price. One production server, one backup/test instance, unlimited internal users, and 12 months of updates and support.

Book a technical fit call
Includes
  • One production server
  • One backup/test instance
  • Unlimited internal users
  • 12 months of updates
  • Onboarding session
  • Direct support channel
  • Compatibility releases
  • Documentation & runbooks

Hardware (Mac mini or Mac Studio) is purchased by the customer. We help size it on the call. After year one, renewal is optional — the installed version keeps running.

Architecture

No cloud inference

Your prompts never reach an upstream model provider. Inference is pinned to your hardware.

Identity

Per-tool API keys

Issue keys to teams and applications. Revoke individually without breaking other consumers.

Logging

Off by default

Prompt contents are not stored. Choose privacy, hash, or audit modes — admin's call, not ours.

Network

LAN & VPN only

Designed for private networking. We do not recommend direct public internet exposure.

Frequently asked

The questions every admin asks first.

No. Outcry is the free App Store app for individuals. Outcry Server is a paid, signed, notarized direct download — so we can support organizational licensing, installers, deferred updates, and direct support contracts.
No. The server detects new Outcry model releases, verifies them against our publisher signature, and stages them in a non-production slot. Production stays pinned to whichever version your admin promoted. You decide when — and whether — to update.
Yes — through VPN or Tailscale-style private networking. We do not recommend exposing the server directly to the public internet, and we'll help you set up a private overlay network during onboarding.
Unlimited internal users. Real concurrent performance depends on your hardware and which model size you run — we'll size it together on the call. A Mac Studio with M2 Max comfortably serves a 30-person organization.
The installed version keeps running. Forever. Renew maintenance for new updates and continued support, or buy a lapsed upgrade later if you skip a year.
Prompt logging is off by default. Admins can choose three modes: privacy (metadata only), hash (one-way hashed prompts for abuse-detection), or audit (full plaintext, with a clear in-product warning). Your choice — not ours.
You need an admin owner and a Mac. The product is designed as a GUI appliance with diagnostics and support included. If you have an IT person, great — they can hand it off as a managed service. If you don't, we onboard your admin owner directly.
Request access

Tell us about your org and your threat model.

This is not a checkout. It's a short form so we can size hardware, scope support, and decide together if Outcry Server is the right fit. We follow up within two business days.

  • No automated marketing. A human reads your form.
  • We'll size hardware and support together on the call.
  • If we're not the right fit, we'll say so.
Your strategy is not for sale.
Required
Valid email required
Required
We use this to follow up — nothing else. No newsletter, no CRM lifecycle.