The platform behind every AI operating system.

An AI operating system runs a firm’s expert work end to end. AI specialists do the groundwork from your playbooks, your people approve the exact version, and clients request, approve and pay in one portal. AgencyOS, LegalOS and every Custom OS run on Stackmint.

See how it works

How an AI operating system works

Governance layerGatewayAudit logsBudgetsKill switches
  1. 04

    Client portal and billing

    Clients send requests, approve deliverables and pay in your branded portal.

    RequestsQuotesApprovalsInvoices
  2. 03

    Workflows

    Specialists hand work to each other and stop at approval gates where your people decide.

    DispatcherSpecialistsYour expertYour client
  3. 02

    AI specialists

    Each specialist does one job well, with a pinned model and declared inputs and outputs.

    ResearchDraftingReviewAnalytics
  4. 01

    Your knowledge

    Playbooks, templates, past work and connected tools become the context every specialist works from.

    PlaybooksTemplatesPast workYour tools

Under the hood, each specialist is an agent that runs versioned workflows built from typed steps, with approval gates where your people decide. Read the architecture

Not a copilot. Not an automation tool. An operating system.

Copilots help one person with one prompt. Automation tools move data on fixed rules. Agent builders leave you to assemble and supervise the agents. An AI operating system runs the whole service, with your people in charge.

CopilotsAutomation toolsAgent buildersAI operating system
Does the work end to endCopilotsOne prompt at a timeAutomation toolsFixed rules onlyAgent buildersIf you build and supervise itAI operating systemAcross specialists, from request to invoice
Follows your methodCopilotsWhen you paste it inAutomation toolsHard-coded stepsAgent buildersWritten into promptsAI operating systemVersioned playbooks
Human approvalCopilotsUp to each userAutomation toolsBolted onAgent buildersYou build itAI operating systemA named person approves the exact version
Client portal and billingCopilotsNoAutomation toolsNoAgent buildersNoAI operating systemBuilt in
Audit, budgets, kill switchesCopilotsChat historyAutomation toolsRun logsAgent buildersVariesAI operating systemOn every run

Every run is visible, bounded and approved.

A client request moving through an operating system. Each step records who or what did the work, with which version, and stays within its budget.

Example runmarket_review · run 0412
  1. Request

    Client requests a quarterly market review in the portal.

    source: client portal · quote approved
  2. Routed

    The dispatcher assigns the research and analytics specialists.

    playbook: market-review v4
  3. Drafted

    Specialists draft the review from your playbook and connected data.

    model pinned · run budget enforced
  4. Paused for approval

    Your expert edits two sections and approves the exact version.

    approver: named strategist
  5. Delivered

    The client approves in the portal. Invoice issued, audit trail sealed.

    every step replayable
Approved by a named personBudget enforcedLogged and replayable

One governance layer under every OS.

Built for firms that answer to clients and regulators. Every operating system inherits the same controls, and you can show exactly how the work was done.

Human approval before anything reaches a client

Nothing reaches a client until a named person approves the exact version.

Immutable, granular audit logs

Every request, decision, change and result is recorded. Answer any client or regulator with evidence.

Kill switches

Stop any agent or workflow at once.

Execution gateway

Agent actions run through a controlled gateway, with credentials kept server-side.

Versioned agents

Each agent, playbook and model version is pinned. No silent changes to how work gets done.

Client data isolation

Every client and matter has its own context. Documents are treated as content, never as instructions.

EU data residency on requestDPA on requestSecurity at Stackmint

Build the operating system your firm runs on.

AgencyOS and LegalOS are ready-made. For any other expert service, we build your OS with you from the same building blocks, around your playbooks and your clients.

AgencyOSLegalOS
  • Agents from your playbooks

    Describe the work in plain language. Stackmint turns it into specialists that follow your method.

  • Governed like software

    Versioned, tested and approved before release, with budgets and policies per team.

  • Delivered where work happens

    Run them in your portal, in Slack, through the API or as MCP tools for your AI assistants.

Start with one team, on real work.

  1. 01

    Working session

    We map your services and playbooks to an operating system, using your real work.

  2. 02

    Pilot on live work

    Start with one team. Your people review every output while the agents learn your standards.

  3. 03

    Roll out across the firm

    Add teams, clients and your own agents as the results come in.