AI made individuals faster.
Maskin makes companies smarter.

Maskin is one shared workspace for your team and its AI agents — where customer signals become decisions, decisions become shipped work, and every outcome makes the next call smarter.

Insight
Onboarding drop-off spike — week 2
Product Analyst
Bet
Onboarding quiz reduces churn 15%
Bet Strategist
Task
Design quiz flow + UI screens
Designer
Developer
Knowledge
Shipped & learned — drop-off ↓18%
All agents
Outcome measured · memory updated
Agent spotted this
Human reviewed & approved
Agents ship it
Outcome measured & learned
insight → bet → task → knowledge → repeat
Apache 2.0 Open source — every line inspectable
MCP-native Built for the agentic era
Bring your own model Anthropic · OpenAI · Ollama
Self-host or managed EU & US data residency
See it in action
Watch Maskin run a full workflow From Slack message to shipped bet — under 3 minutes

Want to see a live walkthrough instead? Book a 20-minute demo with the founders →

Don't just watch it.
Run it on your own problem.

Describe a problem, paste notes, or drop a doc — Maskin drafts a scoped bet back in seconds.

Or set up your first loops:
Today

Faster output.
No better outcomes.

  • Your team ships more — but you're not sure you're building the right things.
  • AI insights live in private chats. No one else sees them. Nothing compounds.
  • Decisions happen in DMs, get lost in Slack, and start from scratch next sprint.
  • The company invests in AI. The P&L doesn't know.
With Maskin

The right bets,
running themselves.

  • Agents surface signals. Humans decide what matters. Work starts automatically.
  • One shared system — every decision, insight, and outcome is visible to the team.
  • Memory persists across sprints. Your team gets smarter every cycle.
  • The loop closes: customer signal → shipped bet → measured result.

The workflows that turn AI spend
into business outcomes.

Pre-built closed loops — each one wires the right agents, objects, and handoffs so your team stops reacting and starts compounding.

Competitive intelligence

Know what competitors shipped before your sales team does

Maskin watches competitor changelogs, pricing pages, and job postings. Filters noise from signal, routes by team — product gets the feature gaps, sales gets the battlecard update.

Intel Monitor Bet Strategist
via web / GitHub changelogs Signal Battlecard / roadmap bet
Lost deals → roadmap

Turn every lost deal into a roadmap bet

Every closed-lost reason in HubSpot gets synthesised into product insights, weighted by ARR. The Bet Strategist surfaces the patterns your roadmap is missing — and closes the loop to sales when they ship.

Deal Analyst Bet Strategist
via HubSpot Loss patterns Roadmap bet
Churn & expansion

Catch at-risk accounts before they churn

Usage drops, login gaps, error spikes — the Churn Monitor watches the signals that matter. When a risk pattern fires, the right person gets an alert with full context, not a dashboard to check.

Churn Monitor CS Advisor
via PostHog / HubSpot Risk signal CSM intervention
Customer feedback

From customer complaint to shipped fix — automatically

A customer posts feedback. It becomes an insight, gets shaped into a bet, built and shipped — and when it goes live, the customer who asked for it is notified automatically.

Feedback Monitor Bet Strategist Developer
via Slack / Intercom Insight Bet Shipped → Customer notified
Market signals

Turn market noise into prioritised bets

Maskin monitors Reddit, X, LinkedIn and community forums. Extracts what's relevant, synthesises it into insights, and surfaces the bets worth acting on — before your competitors do.

Signal Monitor Insight Synthesiser Bet Strategist
via Reddit / X / LinkedIn Insights Bets
Custom Build your own

Any workflow your team runs

Compose your own closed loop. Add agent roles, wire the objects, define the handoffs. If your team runs it today, Maskin can run it with you tomorrow.

Your agents Your objects

Signal in. Decision made.
Work done. Loop closed.

Four moving parts. Every workflow in the marketplace runs on the same primitives — wire them once, they run forever.

01

Triggers

Work enters the system automatically — a Slack message, a new complaint, a morning scan. No one has to remember to kick it off.

Slack message → new insight
02

Agents

Agents have defined roles and stay on the job. The Feedback Monitor reads signals. The Bet Strategist shapes them. The Developer ships. Your team stops being the glue.

Feedback Monitor Bet Strategist Developer
03

Threads

Every decision has a thread. Humans and agents post to the same one. The full context — why a bet was made, who approved it, what shipped — stays attached to the work. Not buried in Slack.

Agent: "Riskiest assumption flagged ↗" PM: "Approved — ship it"
04

Integrations & MCP tools

Agents connect to the tools your team already uses — Slack, GitHub, Linear, Intercom, PostHog, HubSpot and more. No new stack. No bespoke connectors. Just agents that know how to use what you have.

Slack GitHub Linear + more
"A team of three humans and four agents can outperform a team of twenty. The constraint was never speed."

AI made everyone faster. Few teams got better. The bottleneck was never writing the code — it was deciding what to build, and knowing whether it worked.

Maskin is built for that constraint. Humans and agents share one system, one memory, one set of decisions — a signal surfaces, a human shapes the bet, agents ship it, the outcome is measured, and the next bet starts smarter. Judgment still decides which bet is worth making. That's the part that doesn't delegate.

Your AI investment,
finally visible.

Decisions that compound
Every insight, bet, and outcome is stored — not lost in a chat history. Your team gets smarter each sprint. The agents start the next cycle with everything the last one learned.
You stay in control
Every agent action is visible in the shared workspace. Humans approve bets before they go active. Nothing ships without a decision trail — and every decision is yours to revisit or reverse.
Open & no lock-in
Apache 2.0. Self-host or managed. Bring your own model key — Anthropic, OpenAI, or local via Ollama. No markup on AI spend. No proprietary data format you can't export.

Your data, on your terms.

We're an open, early-stage company — so our security model is built on transparency and control, not a logo you have to take our word for.

Open and inspectable
Apache 2.0 — every line of the platform is public. No black box, no hidden data paths. Audit exactly what runs before you trust it with anything.
Your prompts never become our data
Bring your own model key and your prompts and context go straight to your provider — not through us. Nothing is ever used to train a model. Hosted tenants are isolated, with EU & US data residency.
SOC 2 — on our roadmap
We're working toward SOC 2 Type II. Until it lands, self-hosting gives you complete control of where data lives, and the open codebase lets your team verify our posture directly. Need specifics? Ask us anything.

The questions worth asking
before you sign up.

Will this actually improve business outcomes, or just make the team feel busier?
That's the right question — and the one most AI tools can't answer. Maskin is designed around the outcome loop: every bet has a stated goal, every shipped change is linked back to the signal that triggered it, and your team accumulates institutional memory that makes the next cycle smarter. Whether it improves outcomes depends on whether you're running it on the right workflows. That's what the 14-day trial is for.
How is this different from Claude Projects or ChatGPT Teams?
Those share prompts. Maskin shares state. The insight an agent spotted overnight is the same insight you open at 9am, linked to the bet you shaped yesterday, linked to the task your designer just picked up. A chat history is not a team memory — and a team memory is not a product process.
What happens when an agent does something wrong?
Every agent action is posted as a comment on the relevant object — visible to the whole team. Humans approve bets before they become active. Nothing deploys without a human in the loop. If something looks wrong, you close the bet and the agents stop. The audit trail stays.
Do I need engineers to run this?
No for hosted — sign up, invite the team, plug in a model key, go. Yes if you want to self-host, but it's one docker-compose file away.
What's a "bet"?
A scoped hypothesis with a goal, evidence, and a timeline. Borrowed from Shape Up. It's the unit where leadership judgment compounds — signals get surfaced, priorities get chosen, outcomes get tracked. Bets replace the backlog.
What models can I use?
Bring your own. Anthropic, OpenAI, local via Ollama — anything that speaks a standard API. Your key, your spend, your rate limits. No markup.
Where does my data live? Does it train anything?
Hosted: isolated tenants in EU or US regions. Self-host: wherever you put the container. Either way your prompts and context are never used as training data — they belong to you.
What does it cost?
14-day free trial — no credit card needed. After that: $20/seat/month (Pro), $200/workspace/month flat (Team, unlimited seats), or Enterprise with your own LLM and infrastructure. Self-host is always free under Apache 2.0.

Start free. Scale when
it's earning its keep.

14-day free trial. No credit card. Bring your own model key — no markup on AI spend. Self-host is always free under Apache 2.0.

Trial
Free/14 days

Full product. No credit card. Enough usage to run real workflows.

  • 1 workspace
  • Up to 5 members
  • Full marketplace access
  • Generous trial usage included
  • Community support
Start free trial
Team
$200/workspace/month

Unlimited seats. One flat rate. Everything included.

  • Everything in Pro
  • Unlimited seats — flat price
  • $200 in tokens included / month
  • Self-host option
  • Bring your own model key
  • Dedicated onboarding
  • SLA & invoicing
Talk to us
Enterprise
BYOL

Bring your own LLM. Run on your own infrastructure. Full control.

  • Everything in Team
  • Bring your own model — any provider
  • Self-host on your infra
  • SSO & RBAC
  • Custom data residency
  • Dedicated support & SLA
Contact us

14-day free trial · no credit card required. Self-host is always free — Apache 2.0.
Included tokens cover everyday usage out of the box; connect your own model key for heavier workloads — no markup, your spend, your rate limits.

Your competitors are already
running closed loops.

We're onboarding design partner teams now. Get a live walkthrough with the founders — or clone the repo and run it yourself today.

Prefer email? ai@maskin.io

Built in Copenhagen by Værksted.