Refix

Refix es un analista de IA para los datos de la empresa. Haz preguntas sobre el negocio, analiza las fuentes conectadas, configura alertas y ponte al día con los hallazgos de tu asistente de IA.

Servidor MCP alojado

npx add-mcp 'https://mcp.refix.ai/mcp'

Se instala en Claude Code, Codex, Cursor y más

Documentación

Example goal More completed bookings

01

Spots what's worth investigating.

Refix APP

Searches are rising, but bookings are flat. I started looking into it. Want me to keep going?

demand is here bookings never follow Searches 186 Bookings 94 Index

David 9:42 AM

Yes. Find out why.

Investigation continues

Example goal More completed bookings

02

Asks the person who knows.

Refix APP 11:18 AM

These slots say “available.” Can customers book them online?

Claire 11:20 AM

No. Those slots require a phone call.

Answer added to the investigation

Example goal More completed bookings

03

Brings a test worth running.

Claire’s answer changes the test.

Read Claire’s reply Some slots require a phone call

Saved the booking rule Available to future work

Show instantly bookable slots first.

Approved by David Growth

Example goal More completed bookings

04

Comes back with the result.

Refix APP 4:32 PM

The test increased completed bookings. Keep the change.

TEST LIVE bookable slots first +71 pts vs the stall Searches 185 Bookings 171 Index

David 4:34 PM

Good. Keep it running.

Give Refix a goal to see through.

Choose an outcome, from repeat purchases to revenue in a new market. Refix pursues it over days and weeks, using each finding to decide what to investigate next. The work continues between your conversations.

Improve onboarding conversion

Find where new users stall, which groups are most affected, and what changed.

View original### Increase repeat purchases

Find what brings customers back for a second order, and what gets in their way.

View original### Reduce customer acquisition cost

Trace acquisition through activation and payment to find where spend fails to produce customers.

View original### Reduce subscription churn

Understand why subscribers leave and where your team can give them a reason to stay.

View original### Increase trial-to-paid conversion

Find what separates paying customers from users who activate but never buy.

View original### Grow customer referrals

Follow invitations through to new customers. Find where a recommendation loses momentum.

View original### Increase average order value

Find useful bundles and recommendations that grow each order while protecting margin.

View original### Increase search-to-booking conversion

Investigate where searches fail to become bookings, from available inventory to checkout and payment.

View original### Reactivate inactive customers

Find why customers stopped coming back and what has changed enough to bring them back.

View original### Grow revenue in new markets

Spot demand you aren't serving well and find which markets deserve your next investment.

View original

Refix keeps what it learns. So the next answer takes less work.

How quickly Refix finds something useful, how often teams act on it, and what gets followed through.

60%

Proactive findings that led to action

  • Acted on 30
  • Reviewed, no action 15
  • Awaiting review 5

30 of 50 findings led to action.

18 hours

To the first useful finding

0 hours 18 hours

Median time from connection to a finding the team confirmed was useful.

90%

Experiments reviewed without a reminder

  • Results assessed 18
  • Ready for review 2

18 of 20 experiments with enough data received a results review.

Tested on hard data questions.

DataAgentBench, from UC Berkeley's EPIC Data Lab and Hasura PromptQL, asks AI agents real analysis questions across messy, multi-source datasets.

Read how we tested

88.9% reviewed coverage

Questions answered correctly

48 of 54

Datasets, from reviews to patents and sales

12

Preliminary results. Official results publishing soon.

Read-only by design

Writes are refused at the database, not left to the model.

Approvals you control

Require your team's approval before Refix takes action in connected tools.

Roles for every teammate

Decide who connects data, approves actions, and manages billing.

Usage you can see

Daily credit use, with an alert before you reach your limit.

Zero data retention

Model providers keep nothing and never train on your data.

Try it free for 14 days.

The full agent on every plan.

Pro · 14-day free trial

$150 $1,500 per month per year

Start with 14 days free and 70,000 credits. Then capacity for ongoing investigations, with a product manager sitting with you.

  • 150,000 credits / mo 1,800,000 credits / yr
  • All features included
  • Forward deployed PM

Enterprise

Custom pricing

Run Refix in your cloud. Air-gapped option, your models, private access.

  • Everything in Pro
  • Air-gapped self-host on Kubernetes
  • Private dashboard and OIDC
  • Your LLM gateway
  • Restricted outbound connections
  • Dedicated onboarding

Before you get started.

How Refix fits into your tools and your team.

Talk to a human

How does Refix work with our existing analytics?

Keep the tools you already use. Refix connects to their data to investigate what's happening across your business and where there's room to grow.

What makes Refix proactive?

Refix chooses what to investigate from your connected data and business context. It carries the work forward between conversations and returns when there's something useful to act on.

Do we need to know what to ask?

No. Refix looks for opportunities itself. Bring a question when you have one, or give it a goal to pursue over time.

How do Home and Slack work together?

Home brings together your brief and investigations. Slack lets Refix work with your team in conversation, and email brings updates back to you. You can get started without Slack.

What can Refix change?

Refix uses read-only access for analytics and database investigations. Actions in connected tools follow your workspace's approval settings.

How soon can Refix begin useful work?

Connect a data source to start your first investigation. Refix builds its understanding of your business as it works.

Can Refix run in our cloud?

Yes. Run Refix on your own Kubernetes cluster, using your own data services and LLM gateway.

How does pricing work?

Pro costs $150 a month or $1,500 a year, with a 14-day free trial. Plans differ in credits and support. Enterprise includes self-hosting and help deploying Refix.