home →
Get the free key
v1.x — live & shipping weekly

Monthly plans

The free tier is a real product, not a trial — the key is published below and never expires. Paid tiers bill monthly, with no annual option and no lock-in. Every tier keeps your memory, your files and your dashboard on your own machine.

Free · forever
LIVE
now

Developer

$0
🔌1 backend
free for life on v1.x · no account
The whole memory engine, permanently free. Not a trial, not a demo, no card — the key below is the same one everybody uses.
15 MB
per file
∞
users
Your key — no sign-up needed
4XC7N-K2RS3-NXSF6-M5CXA-AL8PU

Paste it at first boot, or into the dashboard. Works offline forever.

  • The memory engine, in full
  • ✓ Unlimited memories, files and users — no counters, no quotas LIVE
  • ✓ Three-lane fact extraction — 320 regex extractors + spaCy, plus a local-LLM pass
  • ✓ Hybrid retrieval — keyword and semantic search fused with RRF
  • ✓ 10 memory categories with priority, always-inject and expiry
  • ✓ The Gardener — idle-time dedupe and contradiction cleanup
  • ✓ Memory chip on every reply — see it, rate it, forget it
  • Files & documents
  • ✓ Upload 33+ file formats and just ask — 13 Inspector analyzers build the dossier
  • ✓ Chunked resumable uploads, per-user folders, chunk-level citations
  • Users & safety
  • ✓ Complete per-user isolation — five-tier identity resolution
  • ✓ Token budgets, curfews and parental controls with a server-verified PIN
  • ✓ Password-protected dashboard (PBKDF2) even from localhost
  • Surfaces
  • ✓ Admin dashboard — Easy and Expert modes, 500+ settings, live apply
  • ✓ Browser chat, terminal client, uptrim-cli and desktop window
  • ✓ Works with Open WebUI, SillyTavern and any OpenAI-compatible client
Get the free key →
Standard
web
aware

Standard

$5
🔌5 backends
per month · the one most people want
Your AI stops being stuck in its training data. Ghost researches the live web inside the time your request was already going to take.
25 MB
per file
∞
users
  • ↓ Everything in Developer, plus: ↓
  • Ghost — live web research
  • ✓ Ghost research engine — specialists race a 5,000 ms window, zero added latency LIVE
  • ✓ 13 specialists scored from 15 signals; 6 keyless search backend families, pluggable
  • ✓ Answer distillation into a short sourced brief, plus an honest source status line
  • ✓ Research mailbox — good findings carry into the next relevant turn
  • ✓ Reflex search — catches "I don't have current information" mid-stream ONE SWITCH AWAY
  • Integrations
  • ✓ MCP server — 40 tools including 10 LSP tools, for n8n and Claude Desktop LIVE
  • ✓ Reads auto-allowed, writes gated behind dashboard approval
  • ✓ Remote access — one-click Cloudflare Tunnel or Tailscale, with a watchdog
  • Routing & agents
  • ✓ 5 backends in one dropdown · model aliases · auto-routing by task complexity
  • ✓ Custom system prompts per user and per model
  • ✓ Agent mode — 17 tools, hard caps, fail-closed sandbox ONE SWITCH AWAY
Join the waitlist →
Most Popular
Pro

Pro

$15
🔌8 backends
per month · local and cloud, one memory
Draft on your own hardware, escalate to a frontier model when it's worth it — both sides see the same memory. Local stays the default.
1 GB
per file
∞
users
  • ↓ Everything in Standard, plus: ↓
  • Cloud, on your terms
  • ✓ Four cloud providers — Anthropic (OAuth, no API key), OpenAI, Gemini, OpenRouter's 100+ models ONE SWITCH AWAY
  • ✓ local_only mode hard-disables all of it — nothing leaves the machine
  • ✓ Cost ledger — every cloud call logged with its estimated cost
  • Orchestration
  • ✓ The Conductor — splits a turn into typed subtasks, routes each to the best backend under cost and latency constraints EXPERIMENTAL
  • ✓ Agent swarm — sub-agents run subtasks in parallel ONE SWITCH AWAY
  • ✓ Verified swarm research — skeptic judges refute by default, only confirmed findings return ONE SWITCH AWAY
  • ✓ Ghost Mesh — an analyst that mines the knowledge graph for gaps and a predictor that pre-warms the search cache, on local heuristics with no LLM call ONE SWITCH AWAY EXPERIMENTAL
  • Room to grow
  • ✓ 8 backends · 1 GB per file · burst pool for deep queues
  • ✓ Generation scheduler — real queues, priority classes, visible "queued #2 · ~40 s" ONE SWITCH AWAY
Join the waitlist →
Premium
👑
deep
brain

Premium

$30
🔌10 backends
per month · for people who live in it
The layers that only pay off after months of use: a graph of how your facts relate, and a proxy that consolidates them while you sleep.
5 GB
per file
∞
users
  • ↓ Everything in Pro, plus: ↓
  • The deep memory layers
  • ✓ Knowledge graph — entities and relations in a temporal graph; a 2-hop walk surfaces memories neither search would find LIVE
  • ✓ REM consolidation — while you're away it merges duplicates, links distant facts and forms hypotheses LIVE
  • ✓ Overnight digest — "last night I connected X to Y" waiting on /my-memory
  • ✓ Predictive context — pre-loads what you'll probably need ONE SWITCH AWAY
  • Watching it work
  • ✓ TUI monitor — boot-time terminal dashboard: live status, backends, users, logs
  • ✓ Belief Inspector — ask why the proxy believes something and get the evidence trail
  • ✓ Dev settings · 10 backends · 5 GB per file
  • On the bench, honestly labeled
  • ✓ Atlas belief engine and Muse injection brain — shipping, measuring themselves against the current path before they take over SHADOW
  • ✓ Mind Mode — 2–3 of your models reason together; never fires unless you ask for the mind model EXPERIMENTAL
Join the waitlist →
Business
team
ready

Business

$100
🔌15 backends
per month · up to 40 users
Several proxies, one memory. Set up with us rather than through a checkout — fleet deployments need a conversation first.
100 GB
per file
40
users
  • ↓ Everything in Premium, plus: ↓
  • Multi-node fleet
  • ✓ One memory across many machines — the master owns the database, workers funnel every read and write to it ONE SWITCH AWAY
  • ✓ Workers boot keyless and take an HMAC-signed grant from the master — never a key on a worker
  • ✓ Master outage degrades to stale-serve, then an honest failure — never a silent fork
  • ✓ Honest notes: one build across the whole fleet, and TLS is yours to terminate (custom CA supported)
  • Running a team on it
  • ✓ Centralized admin · users and settings replicate down · live role flip from the dashboard
  • ✓ SSO / bring-your-own auth · team memory pools with roles
  • ✓ Team analytics · audit log · SIEM export
  • ✓ Scale profiles size ten interacting caps in one move — 40, 150 or 400 concurrent
  • Support
  • ✓ Answers within one business day · onboarding and migration help
Talk to us →
Enterprise · custom
let's
talk

Enterprise

Let's talk.
🔌∞ backends
quoted · org-wide deployments
On-prem licensing, federation across regions, your own key custody, and data-residency rules written to your requirements.
no cap
per file
unlimited
users
  • ↓ Everything in Business, plus: ↓
  • Deployment
  • ✓ HA failover across nodes
  • ✓ On-prem licensing — air-gapped deployments supported
  • ✓ Multi-proxy federation across regions
  • ✓ Unlimited backends, users and file sizes
  • Governance
  • ✓ BYOK / KMS key custody
  • ✓ Custom data-residency rules
  • ✓ Compliance pack · custom retention and deletion policies
  • Partnership
  • ✓ Dedicated support · a direct line to the person who wrote it
  • ✓ Roadmap input and prioritized requests
Talk to us →
$0
The free tier is the product
Unlimited memories, files and users on the Developer key. No counters, no per-month upload quota, no expiry date.
14 days
Trial, and offline grace
A 14-day free trial at first boot needs internet exactly once. After that, licenses keep validating offline for up to 14 days.
Monthly
No annual option, no lock-in
Cancel and the proxy drops back to free Developer features. Your memories, files and settings stay on your disk either way.
🚀

Upgrading takes two minutes

Start on the free key. Move up only when you actually hit the ceiling.

1

Start free

Download the single binary, run ./ContextProxy, paste the free key. Point your chat UI at http://<host>:9099/v1 and carry on as normal. No account, no card, no Python.

2

Hit the ceiling

Usually it's the second LLM box, a file bigger than 15 MB, or wanting your AI to know what happened this week. Pick the tier that removes the thing actually blocking you.

3

Paste and go

Drop the new key into the dashboard. Nearly every setting applies live, so features unlock on save without a restart. Stop paying later and it falls back to the free key — nothing is deleted.

License FAQ

Keys, trials, and what happens when you stop paying.

Is the free tier really free, or is it a trial?
Really free. Developer is $0 forever on v1.x and the key is published right on this page — 4XC7N-K2RS3-NXSF6-M5CXA-AL8PU — the same key for everyone, no account, no card, no expiry. It gives you the whole memory engine with unlimited memories, unlimited files and unlimited users; one backend and 15 MB per file are the only caps. There's a separate 14-day free trial if you want to test the paid features before deciding; it needs internet exactly once to start, and it's the only time-boxed thing here.
How does monthly billing work?
Pick a tier, pay monthly, use it. There is no annual plan and no multi-year contract — we removed annual pricing entirely rather than pretend a discount for locking you in is a favor. Your license activates immediately and renews on the same day next month. Cancel whenever; you keep full tier access to the end of the cycle you already paid for. Refund terms are on the legal page.
What happens if I cancel, or my card fails?
The proxy drops back to free Developer features and keeps running. Your memories, files, settings and uploads are on your own disk — nothing is deleted, nothing is locked, nothing is held hostage. Re-activate later and the paid features come back on save. You would lose access to the paid layers (extra backends, Ghost, cloud providers, the knowledge graph) but not to a single thing you've stored.
Does it need to phone home? What if I'm offline?
Licenses validate offline for up to 14 days, so a flaky connection, a locked-down network or a fortnight on a boat doesn't stop you working. The only moment internet is strictly required is starting the 14-day trial, and activating or changing a key. The free Developer key works offline indefinitely. None of this is telemetry — there is no usage reporting, and there are no analytics on this website either.
What actually changes between tiers?
Three ladders and a feature set. Backends: 1 → 5 → 8 → 10 → 15 → unlimited. Max file size: 15 MB → 25 MB → 1 GB → 5 GB → 100 GB → unlimited. Users: unlimited everywhere except Business, where the license is written for up to 40.

Feature-wise: Standard adds Ghost live-web research, the MCP server, tunnels, aliases, auto-routing, custom prompts and agent file tools. Pro adds the four cloud providers, the Conductor, agent swarm and the cost ledger. Premium adds the knowledge graph, REM consolidation, predictive context and the TUI monitor. Business adds multi-node fleet, centralized admin, SSO, team analytics, audit log and SIEM export. There are no per-month upload quotas at any tier.
Why can't I just buy Business or Enterprise here?
Because we'd rather not sell you a fleet through a checkout box. Multi-node deployments need a conversation about the shape of your setup first — master and workers must run the same build, so upgrades are atomic rather than rolling, and transport between nodes is plain HTTP unless you terminate TLS yourself (a custom CA is supported). Both tiers route to contact and we'll size it with you.
Does a paid tier send my data anywhere?
Only if you switch cloud on, and only the specific request you route there. The memory brain — extraction, embeddings, the knowledge graph, overnight consolidation — runs on your hardware against a local database at every tier. Pro's cloud providers are opt-in, and local_only mode hard-disables them. Ghost fetches web pages you asked about, DNS-pinned and SSRF-filtered so redirects and rebinding fail closed. There is no UPtrim cloud holding your conversations, on any tier.
What do I get if I have a problem?
A reply within one business day, from the person who wrote the code. Bugs go on the public GitHub issues tracker where you can watch them get fixed. The support page covers the common failures first — starting with the Open WebUI environment variable, which causes more setup problems than everything else combined.

Start free. Upgrade only if it earns it.

v1.0 is live and v1.x ships weekly. The free key never expires, so the honest way to evaluate this is to run it against your own stack for a month and see whether your AI actually starts knowing who you are.

bored? try typing “pong” or “tetris”

Get the free key See everything in the box →
Got your key? Download & install ↓