Matron — the governed control plane for your agent crew

Autonomous agent teams

A crew of agents is only as strong as the management over it.

Matron is that management layer. It runs your agents as one governed team — modelled on a well-run program-management office, where standards, monitoring, and sign-off are the job, not an afterthought.

Consistent standards

One bar, held everywhere.

Every agent works to the same conventions, seam-contracts, and QA gates. No drift between who built what — quality that's repeatable, not personality-dependent.

Connected & governed

Nothing happens in a silo.

One registry ties every project, repo, and run together; Sentinel watches health and a tamper-evident audit trail records it all — connected monitoring and governance on one surface.

Less token bloat

Spend goes to output, not waste.

Ungoverned agents burn tokens re-reading context, redoing work, and looping. Matron budgets every run, attributes every token to a project, and flags unattributed spend before it compounds.

A PMO, not a prompt

A real division of labour.

Architect plans, engineers build in isolation, QA gates, Forge integrates, Scribe verifies, Sentinel watches. A program-management team — run by one operator.

The crew — and everyone you snap on

A senior team on the night shift. Snap on as many more as the work needs.

Matron ships with a built-in crew. Add an independent specialist, or a whole team in its own repo — every one runs the same iron rule: reads are free, writes are proposals.

Matron's crewbuilt in, on staff from day one

scout

R&D Scout

Researches improvements, new APIs, and competitor moves for what you're building.

readsthe web, your repos

proposesbacklog items, briefs

guarded byresearch only, never ships

curator

Curator

Keeps the registry honest — stall detection, reconciliation, stage-gate nominations.

readsrepo activity, registry

proposesstatus changes

guarded bynever decides, only nominates

architect

Architect

Turns "build X" into an ordered, path-scoped task plan for the coding squad.

readsthe codebase, the ask

proposesbuild tasks

guarded byplan-only, writes no code

sentinel

Sentinel

Watches everything live — uptime, errors, deploys, cost — and raises the alarm.

readsmonitoring, deploys

proposesincidents, remediations

guarded bysuggests, never auto-applies

postman

Postman

Triages your inbox per project, bundles what needs you, and drafts the replies.

readsGmail, per-project labels

proposesdraft replies, labels

guarded bydrafts only, never sends

scribe

Scribe

Writes the daily briefing and weekly review — and verifies every proposal against its evidence.

readsevery agent's output

proposesthe briefing, verdicts

guarded bythe last check before you

Snap on a specialistindependent agents, each in its own repo

＋ snap-on · pixelpioneer-mcp

pixelpioneer

PixelPioneer

Turns a storyboard brief into a finished 9:16 ad spot — product stills plus Kling video — from any Claude client.

readsa brief + the product catalog

proposesstills, music, an assembled MP4

guarded byrenders only on an explicit call

＋ snap-on · bulletin

bugle

Bugle

Scans your industry each week, synthesizes the signal with Claude, and emails you a digest and PDF.

readspublic sources, via Tavily

proposesa weekly digest + PDF

guarded bypublic reads, sends a digest, nothing else

＋ snap-on · factoree

factoree

Factoree

A supervisor phones in a Mandarin shift report; Factoree resolves it against Kingdee ERP and files it in minutes.

readsa spoken report + scanned cards

proposesa readback card, then a filed report

guarded bynever saves until confirmed

＋

Bring your own

Any agent is a role with a guardrail. Point Matron at your MCP or repo and snap it in.

Snap on a whole teameach in its own repo, its own colour

devcrew6 agentsrepo · matron

Turns a build request into shipped code — planned, built in isolation, tested, then integrated for your approval.

architect

Turns a build or fix into an ordered, path-scoped task plan.

plans only — writes no code

backend

APIs, services, auth, DB schema, migrations, data pipelines.

own worktree, never main

frontend

Components, pages, client state, styling, accessibility.

own worktree, never main

flow

n8n workflows, webhooks, nodes, the email plumbing.

go-live stays a proposal

forge

Merges the branches in order and drafts the PR body.

never opens the PR or deploys

Authors e2e tests, runs the suite, calls ship / no-ship.

read + test only

Every engineer builds in its own isolated worktree. Push, merge, and deploy stay your proposals.

opscrew8 agentsrepo · opscrew

A bilingual crew for IT operations: voice and text helpdesk, alert correlation, incident briefs, and guarded remediation.

Front Desk

Bilingual (EN / 中文) voice agent: verifies ID, checks incidents, files tickets.

MFA before any account action

Text Desk

The same helpdesk over MCP for Claude, Cursor, and agentUI.

tenant-scoped by API key

Correlator

Ingests Zabbix / Prometheus / ELK webhooks; clusters and dedupes alerts.

read-only ingest, never acts

Incident Analyst

Writes incident briefs with ranked causes; drafts the post-mortem.

drafts only, never remediates

Remediator

Proposes runbook commands from the knowledge base for approval.

engineer approves; deny-list blocks destructive

Problem Manager

Detects recurring failures, opens problem records, writes the RCA.

hypotheses only, no fixes

Shift Scribe

One-click shift-handover digest for follow-the-sun teams.

summarizes only

Account Manager

Monthly service report: SLA attainment, MTTR, top incidents.

read-only reporting

Atlas7 agentsrepo · careerproof / atlas

Turns raw CVs and a job description into ranked, evidence-backed shortlists — parse, score, interview, report.

Intake

Parses CVs into structured profiles; reformats to your template.

parses only, never scores

JD Analyst

Analyzes a JD for clarity, bias, market fit, and required skills.

analyzes, writes no policy

Fit & Rank

Scores candidate-to-JD fit and ranks a whole pool.

scores; a human decides

GEM Evaluator

Runs competency and custom-rubric evaluations, singly or in batch.

advisory scores, not a gate

Interview Designer

Generates tailored questions, follow-ups, and dialogue assessments.

drafts; the interviewer decides

Reports & Analytics

Compiles talent reports and pipeline analytics.

reports; decides nothing

Advisor

Conversational TA assistant, grounded in your candidates and JDs.

advises, never acts

Build your own team

Group your agents into a crew with a shared repo and a lead. Matron gives them the registry, the worktrees, and the guardrail — you give them the mission.

The operator console

Everything above, on one on-premise screen.

Spend, crew, builds, secrets and a tamper-evident audit trail — 13 governed pages that ship with every Matron instance. On-prem, auditable, yours.

matron · control plane · overview

Overview

Your whole operation on one screen — portfolio health, pending decisions, spend and live runs, at a glance.

approvals

Approvals

Nothing ships without you. Diff, evidence and one-click decide on every change.

spend & tokens

Spend and tokens with caps, forecasts and unattributed spend flagged

Spend & Tokens

Every token accounted for, no mark-up. Caps, forecasts, and unattributed spend flagged.

crew

Crew

Your agent team, managed — model, budget and permissions, with live status for each.

fleets & runs

Fleets and runs: fleet instances, live run logs and on-prem GPU utilisation

Fleets & Runs

Watch the build happen — fleet instances, live run logs and on-prem GPU utilisation.

projects

Projects

The whole portfolio, stage-gated — registry and kanban from idea to live, with stall detection.

repositories

Repositories

Every repo, aligned — Curator keeps them in sync; worktrees and deploy keys in one view.

monitoring

Monitoring

Know it's healthy. Sentinel uptime, error rates, p95 latency and incidents with remediations.

holohud

HoloHUD

Watch the crew work, live — runs, tool-calls and hand-offs, every gate visible.

audit log

Audit Log

A tamper-evident record of everything — append-only, filterable, every action attributable.

governed pages

100%

actions audited

self-approvals

on-prem surface

Illustrative data — the live console runs on your own on-premise Matron instance.

Pricing

Turn product-manager seats into engineering capacity.

One price per seat, two ways to run it. A seat is a person directing the crew; a build is one Dev Crew run — a fix, a feature slice, a proof-of-concept — consuming ≥ 1M tokens.

Your keys

Run it yourself

S$599 / seat / month

Runs on your own keys and infrastructure — you stay in full control.

The full crew + devcrew
First 4 builds / day / seat included
+S$20 / month per extra build/day
Tokens billed direct to your own provider

Managed

We run it for you

S$699 / seat / month

We host and run the governed environment for you, end to end.

Everything in Your keys
Base platform + oversight per seat
Usage at S$2 / hour-block
Tokens at cost + 8% handling, fully reported

Members' calculator

See exactly what your team saves.

The interactive value calculator is behind a quick email check. Drop your work email and we'll send a magic link — no password to remember.

Your link stays valid for 3 days. We'll add you as a lead and may follow up about Matron — unsubscribe anytime.

Partners & Enterprise

Special rates for SI partners and in-house dev teams.

Running Matron for your clients, or across a large engineering org? We tailor both the commercials and the deployment — volume seats, white-label, private cloud, and bespoke crews. Talk to us and we'll quote against your delivery model.

volume

Volume & reseller pricing

Tiered seat discounts past 25 seats, plus margin for systems integrators who resell Matron to their own clients.

white-label

White-label control planes

Matron under your brand — one isolated stack per client, managed centrally from your partner console.

deployment

Private cloud / on-prem

Deploy inside your own VPC or data centre, including data-residency and behind-the-firewall installs (PDPA / regulated).

custom build

Bespoke agents & crews

The devcrew builds you custom agents and teams wired to your stack — ITSM, ERP, ATS (ServiceNow, Kingdee, PeopleSoft).

governance

SSO, RBAC & audit

OIDC + MFA, role-based access, dual-control policy packs, and exportable audit trails for compliance sign-off.

success

SLA, onboarding & enablement

Priority incident SLAs, a named delivery lead, and enablement for your PMs and engineers.

Talk to us Regional pricing (SG / MY / enterprise) quoted on request.

Your operation, kept in order — while you sleep.

A crew of agents is only as strong as the management over it.

One bar, held everywhere.

Nothing happens in a silo.

Spend goes to output, not waste.

A real division of labour.

A senior team on the night shift. Snap on as many more as the work needs.

R&D Scout

Curator

Architect

Sentinel

Postman

Scribe

PixelPioneer

Bugle

Factoree

Bring your own

Autonomy you can actually sleep through.

Looking is free.

Every outward move is a proposal.

You decide, from anywhere.

Everything above, on one on-premise screen.

Overview

Approvals

Spend & Tokens

Crew

Fleets & Runs

Projects

Repositories

Monitoring

HoloHUD

Audit Log

Turn product-manager seats into engineering capacity.

Run it yourself

We run it for you

See exactly what your team saves.

Special rates for SI partners and in-house dev teams.

Volume & reseller pricing

White-label control planes

Private cloud / on-prem

Bespoke agents & crews

SSO, RBAC & audit

SLA, onboarding & enablement

The bridge in your pocket.

Your operation, in order — while you sleep.

Sentinel