Self-serve AI agents for the systems you already run · In development

AI teammates that plug into your business

Connect the tools you already use — helpdesk, store, CRM, even the old system with no API — and put AI teammates to work on the repetitive jobs. You do three simple things. MatePlug does the technical rest.

MatePlug is in development — early access opens to the waitlist first. No rip-and-replace, and every agent action graded, logged, and reversible.

mateplug · task flowDESIGN MOCK-UP · DEMO FLOW
HelpdeskNew ticket #4821“Where is my order?”MatePlug agentChecking order status…Reading the ticket…Checking order status…Drafting the reply…Reply draft⏸ Awaiting approvalNothing sends itself
FRAME 1 · READTicket arrivesAgent reads the ticket and pulls the customer's history.
FRAME 2 · CHECKLooks it upQueries your store system for the real order status — never guesses.
FRAME 3 · DRAFTDrafts the replyWrites the answer in your tone, with the tracking link attached.
FRAME 4 · APPROVALWaits for youSending is a graded action. A human approves — one tap, phone or desktop.

How this flow will run: an 8–12 second loop — stream dots travel the connectors, the agent node cycles read → check → draft, and the reply waits, held for approval. With reduced motion it rests on the static frame above. Demo flow — no real customer data.

Working, out loud

You see every step it takes

Agents will stream their work as they go — what they read, what they decided, what they're about to do. No black box, no silent actions. If it can't answer, it hands over to a human with the full context attached.

1,284
Tickets handled todayDEMO
38s
Median first responseDEMO
86%
Resolved without a humanDEMO
7
Actions waiting for your approvalDEMO
agent output · support-mate DEMO DATA
ticket #4821 — “Where is my order?” lookup order #W-90312 in your store system found shipped Tue · out for delivery today, 2–5 pm draft “Hi Dana — good news, your order is out for delivery today between 2 and 5 pm. Track it here…” held sending needs approval (level L2) — sent to Maya's phone for a one-tap decision

The first selling point

Connecting is something you can do

Bringing a system in used to mean IT projects and API documentation. With MatePlug you only ever do three human things — the platform figures out the technical rest, per system, automatically.

STEP 1

Pick your systems

Choose what you already run — helpdesk, store, CRM, warehouse, the old desktop system in the corner. Search by the name you call it. MatePlug works out how each one can connect.

STEP 2

Authorize

Sign in and approve access, the way you'd log in yourself. No API keys to hunt down, no field names to map — proposed mappings are shown to you in plain words, with one real record to confirm.

STEP 3

Say what you need

Describe the job in a sentence — “answer customers asking where their order is.” MatePlug suggests the right template and skills. Can't phrase it? Pick from examples instead of facing a blank box.

How connecting works, in detail →

Scenario templates

Start from a job, not a blank canvas

Each template brings its connections, knowledge needs, permission ceiling and a replay test set. When early access opens, you install one, connect, pass the checks, and go live — graded, step by step.

T1 · SUPPORT Q&A

Support answers

Answers customer questions from your docs and history across email, chat widget, Telegram and WhatsApp. Anything it can't answer goes to a human — with context.

T2 · STATUS LOOKUP

“Where is my order?”

Looks up live order, ticket and parcel status in your systems and answers in your wording. Read-only, so it can run fully automatic.

T3 · TRIAGE

Ticket triage

Classifies new tickets, sets priority, routes to the right team and asks for missing details — first as suggestions, then for real once you approve.

T4 · QUOTING

Quotes & pricing

Drafts quotes from your price lists and rules. Sending a quote always needs a human yes — money promises never go fully automatic.

T5 · FOLLOW-UP

Chasing & reminders

Follows up on stalled tickets and unanswered threads on a rhythm you set, and escalates before things go quiet for too long.

+ MORE

Teach it your process

Show the agent a job once — on your own desktop, with Agent Link — and it drafts an SOP you can read, edit and approve. Your know-how, written down.

All templates, with permission ceilings and go-live steps →

Trust, by design

Power is earned in grades, never granted at once

Before any agent may write anything, it passes a readiness check and replays your real history — scoring at least 85 overall and 90 on write actions, with zero mistaken irreversible actions. Then it climbs:

L1 · Read & draft

Looks things up, answers from your knowledge, and prepares drafts. Nothing is sent, nothing is changed.

Unlock: readiness check + replay test passed.

L2 · You approve

Writes are queued as pending actions. A human approves each one — mostly from the phone — before it executes.

Unlock: 7+ days at L1, ≥80% of drafts accepted.

L3 · Automatic

Whitelisted actions run on their own, with sampling review afterwards. Quotes and other money promises stay at L2 — forever.

Unlock: 14+ days at L2, ≥98% approval success, you switch it on.

Every action is logged and reversible. Writes always go through your systems' own screens and checks — never straight into a database. And your recordings are never used to train models.

The full safety model →

Pricing

Simple subscription, usage in Credits

Final pricing is being set — no numbers until they're final. The shape is decided: a plan per team, with agent work metered in Credits you can watch in real time.

Starter

—

First agents, core templates, L1–L2 grading.

Team

—

More agents and systems, Agent Link seats, L3 automation.

Business

—

Higher volume, priority support, advanced audit.

How pricing will work →

AGENT LINK

For the system with no API

Some tools only exist as a desktop program on one PC. Agent Link lives on that machine and works the screen the way your team does — asking before it takes over, with a floating control bar so you can pause it or grab the mouse back at any moment. Taking over is never the same as approving: writes still wait for your yes.

MOBILE APP

Approvals in your pocket

The phone app does three things: chat with your agents, approve or reject pending actions from a push, and see status and alerts. The heavy setup stays on the desktop console — the decisions come to you.

Be first in line when early access opens

MatePlug is being built now. Join the waitlist and we'll invite teams in order — starting with the jobs on this page.

Join the waitlist