Mault
Ship fast. Without the chaos.
Governance for AI-assisted development: Mault sets your project up so the AI coder you already use works inside rules it cannot quietly skip, and runs multi-agent workflows with real isolation between agents.
Install: 7-day free trial, then $19.99/month. See Pricing.
Works with Claude Code and GitHub Copilot in VS Code.
Physics, not policy. Hooks that block, not prompts that suggest.
The Problem
You can point a fleet of AI agents at your codebase. Without guardrails, they:
- Collide — two agents edit the same file, one overwrites the other
- Drift — agents ignore your conventions the moment context runs out
- Escalate — an agent merges its own PR, ships untested code, and moves on
Better prompts don't fix this; agents ignore documentation. Mault enforces at the tool level: hooks that block writes outside a declared zone, gates that stop a merge without green CI, budgets that cap how much one agent can change.
Set up a project in three steps
The Production Readiness panel walks you through setup. Mault works through labels, pull requests and the review agent, so it is visible to everyone working in the repository.
Three steps:
| Step |
What it does |
| 1. Select coding agent |
Connect the AI coder you already use, so Mault can work through it |
| 2. Select repository |
Connect this project to its repository, or create one if it has none |
| 3. Setup Mault and governance |
Set Mault up in this project and turn on the rules that keep your AI coder honest. You see what will be installed and approve it first, and the install checks itself before it reports success. |
Step 3 installs the governance chain for your stack: pre-commit checks that block a commit instead of suggesting a fix, and the runtime hooks your agents run under. Languages: TypeScript, JavaScript, Python, Java (Maven) and SolidJS.
Mault for Business adds CI/CD pipeline setup, test setup, custom rules, governance checks and AI coder rules, plus the admin configuration panel and enterprise reports. They show in the panel as Mault for Business and are not part of the $19.99 plan.
Multi-Agent Workflows
Guided multi-agent workflows with role separation, worktree isolation and pull-based task scheduling. Each role has a minimum-permission allowlist enforced by hooks: an agent can only do what its role permits.
The role chain
PLANNER ──► ORCHESTRATOR ──► WORKER(s) ──► REVIEW ──► TESTER
│ │ │ │ │
plan.md issues + PRs with reviewed, quality
delegation-map mault-ready merged report
UNIVERSAL AGENT — escape hatch for a blocked workflow (mault-escalation).
The Spec Agent and the Compliance Agent also appear in the panel, locked, as Mault for Business roles.
Click a role in the Agent Workflows panel: its settings are written, an agent session opens with the prompt submitted, and a chime tells you when the step is done and which one is next.
| Role |
What it does |
Cannot do |
| Planner |
Builds a plan with you; you approve every decision |
Write code, create issues, modify files outside .mault/ |
| Orchestrator |
Creates atomic issues with zones, budgets and dependency order |
Write code, create PRs, push branches |
| Worker |
Picks up an issue, writes code test-first, opens a PR labelled mault-ready |
Merge PRs, push to main, leave its zone or budget |
| Review Agent |
Reviews every PR and sends failures back to a Worker. By default (human-in-the-loop) a person merges; in Automated mode it merges once every check passes |
Create PRs, write code, merge with failing CI |
| Tester |
Runs the tests and verifies merged work |
Modify source code, install packages |
| Universal Agent |
Unrestricted escalation handler for a blocked workflow; can act as any role |
— |
Isolation — how agents can't interfere
| Layer |
Mechanism |
What it prevents |
| Atomic tasking |
One issue = one branch = one PR |
Scope creep, tangled changes |
| Worktree sharding |
Each worker gets its own git worktree |
File-level collisions between agents |
| Session identity |
One session file per agent process |
State corruption between concurrent sessions |
| Pre-authorised sessions |
Sessions created before tabs open |
Races during agent start-up |
| Pull-based queue |
Workers take the next unblocked issue from a shared pool |
Static assignment bottlenecks |
The orchestrator orders issues with sequence labels (mault-seq-N); workers take seq-0 first, then seq-1 once seq-0 is closed, and so on. The fastest worker gets the next task.
Runtime enforcement
Governance hooks run automatically during managed agent sessions, and agents cannot bypass them:
| Hook |
When |
What it enforces |
| Agent bootstrap |
Session start |
Task boundaries, fix-first warnings, the claim flow |
| Zone boundary |
Before a write |
No file writes outside the declared zone |
| Budget gate |
After a write |
Warns at 80%, blocks at 100% of the file and line budget |
| Role permission |
Before a command |
Workers can't merge, orchestrators can't push, planners can't write code |
| Test gate |
Before a write |
New source needs a test in the right layer of the testing pyramid |
| Size gate |
After a write |
Files and functions over the size limits are blocked |
| Receipt gate |
Session end |
No exit without a structured completion receipt |
| Merge gate |
Before a merge |
No merge without review triage and green CI |
| Sprint gates |
Before a command |
Capacity, scope, labels, activation and completion |
Hooks deploy when you turn multi-agent workflows on, and update with each new version of Mault. Outside managed sessions they stay out of the way, so they are safe alongside normal development.
Pricing
One plan: $19.99/month, everything included (project setup and the governance chain, multi-agent workflows, runtime hooks). Mault needs an active subscription to run.
- 7-day free trial. A card is taken when you subscribe; the first charge is on day 8. Cancel during the trial and you are never charged. One trial per customer.
- Promo codes are accepted at checkout.
- Sales tax is calculated by Stripe where it applies and added to the price.
- Cancel any time from "Mault: Manage Subscription" (Stripe customer portal). Mault keeps working until the end of the period you paid for.
- Failed payment: Mault stops until the card is updated in the portal; it starts again once the payment goes through.
- Refunds cancel the subscription immediately.
- Offline: Mault keeps working without a connection until the end of the paid period.
- One computer per subscription.
Subscribe from the prompt Mault shows on start, or run "Mault: Subscribe". Checkout and the portal are Stripe pages that open in your browser. Refund & Billing Policy
Getting Started
- Install Mault from the VS Code Marketplace
- Subscribe — Mault shows the subscription screen on first start (7 days free), or run "Mault: Subscribe"
- Open the Production Readiness panel — follow the three setup steps
- Turn on multi-agent workflows in the Agent Workflows panel and click a role to start
Requirements & Commands
Requirements:
- VS Code 1.96+, Node.js 20+
- A trusted workspace
- Claude Code or GitHub Copilot as your coding agent
Key commands (Command Palette: Ctrl+Shift+P / Cmd+Shift+P):
| Command |
Description |
Mault: Subscribe |
Start the 7-day trial and subscription |
Mault: Manage Subscription |
Card, invoices and cancellation (Stripe customer portal) |
Mault: Refresh Subscription Status |
Check the subscription again, e.g. after paying |
Mault: Open Mault Panel |
Open Mault's panels |
Mault: Scaffold Multi-Agent Setup |
Write the multi-agent settings and hooks into this project |
Mault: Install pre-commit hooks |
Install the governance chain for your stack |
Mault: Reverify Install |
Re-run the install checks and show what passed |
Mault: Run Project Health Check |
Audit the project and write a report |
Mault: Open Mault Board |
Sprint and pipeline dashboard |
Mault: Reset Sessions |
Clear stuck agent sessions |
Detailed release notes are on the Changelog tab of this listing.
Data & Privacy Disclosure
Mault performs its governance work locally within VS Code. It connects to the internet only for the items listed below.
What connects to the internet:
- Account: on first start the extension registers this computer with Mault's backend using an anonymised machine identifier; no email is required
- Payments: Stripe Checkout and the Stripe customer portal open in your browser; card details go to Stripe only and never reach Mault
- Subscription checks: on start and periodically, the extension asks Mault's backend whether the subscription is active and until when; the last answer is kept locally so Mault works offline until the paid period ends
- Error reporting: Crash diagnostics, no code content
- Usage analytics: Anonymous product telemetry via Mixpanel — see Telemetry below for exactly what is sent and how to switch it off
AI-powered features (optional, user-invoked only):
- Code snippets may be processed to generate semantic embeddings
- Embeddings are stored for search functionality
- Raw code is not stored; embeddings cannot reconstruct source code
What we don't do:
- No user code is stored on external servers
- No data is used for AI model training
- No payment card details are stored by Mault (handled entirely by Stripe)
Telemetry
Mault sends anonymous product analytics to Mixpanel. Mault's analysis runs entirely on your machine: no source code, file contents, file paths, repository names, or diagnostic findings are ever transmitted.
What a telemetry event contains:
- The event name (a command invoked, an onboarding step reached) and its category — feature, journey, funnel, or conversion
- A per-install anonymous identifier supplied by VS Code (
vscode.env.machineId), sent as distinct_id — never an email, username, or account identity
- Extension version, VS Code version, Node version, and OS platform (
win32, darwin, linux)
- A timestamp
Geolocation is disabled on the Mixpanel client, and nothing is collected in a workspace you have not marked trusted.
How to turn it off. Any one of these three is sufficient — they are independent, and telemetry stops if any of them says stop.
| Setting |
Effect |
mault.telemetry.enabled → false |
Mault's dedicated telemetry switch. Turns off product analytics and nothing else. |
telemetry.telemetryLevel → off |
VS Code's own host setting. Mault honours it, so switching telemetry off across the editor switches Mault's off too. Only the level off does this; error and crash concern crash diagnostics, not product analytics. |
mault.usage.enabled → false |
Also stops telemetry, but it is not a telemetry switch — its job is clip usage tracking and limits. Turning it off changes clip-limit behaviour as well, so prefer one of the two above unless you want both effects. |
If you want the narrow switch, use mault.telemetry.enabled. If you manage the whole editor centrally, telemetry.telemetryLevel is enough on its own.
How to verify telemetry is off:
- Set
mault.telemetry.enabled to false, then reload the window (Developer: Reload Window).
- Open Help → Toggle Developer Tools → Network and use Mault normally.
- You will see no requests to
api.mixpanel.com. With any of the three switches off, Mault returns from analytics initialisation before a Mixpanel client is ever constructed — there is no client to disable later.
For full details, see our Privacy Policy.
Important Disclosures
- Backend Connection: This extension connects to Mault's backend service to register this computer and check the subscription.
- Account Required: A Mault account is created automatically on first use.
- Subscription Required: Mault runs only with an active subscription: 7-day free trial, then $19.99/month. See Pricing.
- Links: Privacy Policy | Terms of Service | Refund & Billing Policy
License
The Mault VS Code extension is licensed under the Apache License, Version 2.0.
Use of Mault's cloud services is governed separately by Mault's Terms of Service.
Legal & Support