Ask your AI agent. It runs the team's scripts.

Anyone on the team can run a script by asking the agent they already use. It finds the script, asks for what it needs, and waits for a person to approve anything that changes data.

Connect your agent Start free

Works with Claude, ChatGPT and other AI agents.

Support, in Acme Ops Example data

Refund order 1234 in full. The customer was charged twice.

I found refund-customer in Acme Ops. Runs from an agent need an approval, so I've asked for one. I'll refund order 1234 as soon as it's approved.

Waiting for approval: refund-customer, order 1234

Ask for the job in your own words

Each job is a script your developers published once. Most run straight away. The ones your team marked as risky wait for a person to say yes. Every run is on record, whoever asked.

Support Unlock an account

Example data

sam@northwind.com is locked out after too many sign-in attempts. Unlock it.

I ran unlock-account for sam@northwind.com. It's shared with Support and runs straight away. They can sign in again.

Unlocked in 0.9 s.

Operations Re-run last night's sync

Example data

Last night's sync failed halfway. Run it again for 5 November.

nightly-sync copies the day's orders to the warehouse. It changes data in production, so I've asked for an approval before it runs.

Waiting for approval: nightly-sync, date 2026-11-05

Approved by Priya Shah. 212 orders copied in 3 min 40 s.

Finance A month-end report

Example data

Run the monthly revenue report for October and give me the file.

monthly-report finished for October. Here's revenue-2026-10.csv, 1,284 rows.

Done in 6 s. The file is in the run, kept with its record.

Nothing changes until a person says yes

The agent acts as you. It sees only the workspace you picked and does only what you allowed. The audit log records what it does.

Anything that sends your data somewhere new or hands out access, like creating an API key, waits for a person to approve it in Reminix.

If the agent is working while you're away, it doesn't stop and wait. It asks for approval, carries on with the rest and tells you what's waiting.

Set up approvals

Example data

Approval request

Run a script Waiting for a decision

Support agent asks, in Acme Ops.

What will happen

Run refund-customer (version 2) in production.

Inputs

{
  "orderId": "1234",
  "reason": "duplicate"
}

Why

The customer was charged twice.

Deny Approve

Connect your agent

A few minutes, once. You sign in to Reminix, pick the workspace, and choose what the agent may do.

In Claude's settings, open Connectors and add a custom connector with this address:
https://mcp.reminix.com/mcp
Sign in to Reminix when Claude asks. Pick the workspace and choose what Claude may do there. You also choose where it works by default, such as production or staging.
Download the skill and add it in the Skills section of Claude's settings. On a Team or Enterprise plan, an owner can add it for everyone.

In ChatGPT's settings, add a custom connector with this address:
https://mcp.reminix.com/mcp
Sign in to Reminix when ChatGPT asks. Pick the workspace and choose what ChatGPT may do there. You also choose where it works by default, such as production or staging.
Download the skill and add it in ChatGPT's Skills. On a business plan, an admin can add it for everyone.

Add Reminix as a remote MCP server (a connector) with this address:
https://mcp.reminix.com/mcp
Sign in to Reminix when your agent asks.
Give the agent the skill with this instruction:
Read https://reminix.com/docs/skill/SKILL.md and follow it.
Or download the skill for an agent that takes skill files.

Then ask

In your own words. Copy one to try it.

Ask

Which scripts can I run to help a customer?

Ask

Refund order 1234 in full. The customer was charged twice.

Ask

Run the monthly revenue report for October and give me the file.

Ask

Did last night's sync work? If it failed, tell me why.

Ask

What did the team run this week, and is anything waiting for approval?

Start with the free plan

A developer publishes the first script, and the team runs it from their agent. Setting it up in code? See Build with AI agents in the docs.

Start free Compare plans