AI for teams running software in production
When something goes wrong with your software, Operate helps find out why and suggests a fix. It also checks for slow queries, security problems and unnecessary costs, and answers your team’s questions in Slack or Teams.
Runs on your servers. Reads your data without changing it. Use the AI model you choose.
Maintain an open-source project? Try Operate on your GitHub issues.
An alert says payments are failing. Support asks why a customer was charged twice. A routine check finds a database query getting slower. In each case, Operate looks at the relevant records, logs and code to work out what happened.
When an alert arrivesFind what caused the error and prepare a fix for an engineer to review.
Before someone reports a problemCheck for slow database queries, risky access settings and resources you may not need.
When your team has a questionAsk in Slack or Teams and get an explanation based on what happened in your systems.
Error Monitoring / Datadog · payments-api
When the payment provider takes too long, the service retries immediately. Those repeated attempts use up the available connections and cause more requests to fail.
- retry: immediate+ retry: exponential_backoff+ max_attempts: 3Proposed change. Engineer review required.
Illustrative investigation using Operate’s workflow. Findings depend on your connected systems and available evidence.
Your stack
To explain a problem, Operate needs to see the relevant code, logs and data. Connect those systems using accounts that can read the information without changing it.
GitHub, Datadog, Sentry, PostgreSQL, AWS, Slack, Microsoft Teams and more.
Source control
GitHub · GitLab
Logs
Datadog · CloudWatch · GCP Cloud Logging · SolarWinds
Databases
MongoDB · PostgreSQL · MySQL
Chat
Slack · Microsoft Teams
Error tracking
Sentry · Rollbar · Bugsnag · Honeybadger · Airbrake
APM & tracing
Datadog APM · New Relic · Dynatrace · Grafana · Honeycomb · OpenTelemetry
CI/CD & deploys
GitHub Actions · GitLab CI · Jenkins · CircleCI · Vercel · Argo CD
Config & feature flags
LaunchDarkly · Statsig · Flagsmith · Split · Unleash · AWS AppConfig
Cloud & infrastructure
AWS · Google Cloud · Azure · Kubernetes · Terraform
Queues & streams
Kafka · Amazon SQS · RabbitMQ · Google Pub/Sub · Redis Streams
Incident management
PagerDuty · incident.io · Rootly · Jira Service Mgmt
Issue tracking
Jira · Linear · GitHub Issues · Asana · Notion
Support
Zendesk · Intercom · Freshdesk · Front · Help Scout
Status & docs
Statuspage · IsDown · Confluence · Google Docs · GitBook · SharePoint
Connections let Operate read information, not change it. Choose the systems needed for the work you want it to do.
Start with the agents and connections we provide. If your team uses an internal system, you can add a tool to read it and an agent to investigate it. For example, a payout agent can check your payment system to find out why a transfer is late. Explore agents and tools.
This example shows how to add your own tool and agent on Enterprise.
get_payout_status(payout_id)
Returns:
status
processing_history
provider_responseThis tool reads the payout records it has permission to access. It cannot change them.
Illustrative workspace showing Operate-published and custom capabilities. Creation controls demonstrate the workflow and do not save a configuration.
An engineer figures out why a payment failed, explains it in a chat and gets back to work. A month later, someone else has the same question. Operate turns the findings into verified articles, so the next person can find the explanation instead of repeating the investigation.
The gateway can time out after the provider has accepted the payment. Retrying without a stable idempotency key can create a second charge for the same order.
Check the payment records before retrying. Use the same order-specific idempotency key across attempts.
Operate creates an article from the cause and evidence found during an investigation.
Your team can edit articles, check that they are correct and manage the library as the software changes.
Operate looks up relevant articles and checks whether the same explanation fits what is happening now.
Illustrative knowledge article and management view.
Explore the knowledge libraryThe same payment problem can need two answers: support needs to explain the charge to the customer, while engineering needs to fix the code. Operate helps both teams understand what happened.
Get the cause, the evidence and a suggested code change to review. Explore the solution
Find out why a service is failing, slowing down or costing more than it should. Explore the solution
Explain a customer’s problem without asking an engineer to start from scratch. Explore the solution
Find out why a payout, order or scheduled job has not completed. Explore the solution
Using Operate
Your team can run Operate and review its findings. If you also need engineers to investigate problems and look after your systems, we can take on that work with you.
Trust and control
Operate needs information from your systems to investigate a problem. You choose what it can read, where it runs and which AI model it uses.
Install Operate with Docker. Use your own AI provider or a model running inside your network.
Give it accounts with access only to the information it needs. Its queries are recorded.
Operate suggests a code or configuration change. Your engineers check it and apply it.
Open the records, logs and code used in the investigation.
The report includes explanations that were considered and ruled out.
The steps are recorded so your team can review how the answer was reached.
The team behind Operate
We’re the engineers building Operate. We also build software for other businesses, so we know the work involved in keeping it running.
40+
Engineers and product builders
Jai Jalan, founder
Former Microsoft and Google engineer
Professional backgrounds of Better Software customers.
About Better SoftwareQuestions
Operate helps teams find out why their software is behaving unexpectedly. It checks connected code, logs and data, explains what it found and suggests a fix. It can also regularly check for performance, security and cost problems. Your team can ask questions in Slack or Teams.
Yes, on Enterprise and Managed. A tool lets Operate read information from a system you use. An agent tells it what to investigate and what to report. For example, you could add a tool that reads payout records, then an agent that uses those records to explain a late transfer.
Engineering uses it to investigate failures and review fixes. DevOps uses it to check how systems are running. Support asks why something happened to a customer. Operations asks why a payout, order or job has not finished. Each team can use the same connected information to answer its own questions.
It can be. Think about the questions your team asks engineers, such as why an order failed or why a report is late. Operate can investigate those without an alert. It can also check for problems such as slow queries and unnecessary infrastructure costs.
You can use an AI agent that way. Operate provides agents we build, test and maintain, along with connections to your systems and regular checks. Enterprise also keeps verified articles from earlier investigations and lets you add your own tools and agents. If you need people to handle the work too, our engineers can take on agreed responsibilities through Managed.
Anyone in the Slack or Teams channels where you add it. A support rep could ask why a customer was charged twice; an engineer could ask why requests started failing. The answer comes back in the thread, with links to the information used.
No. Keep using them to collect logs and raise alerts. Operate reads the relevant information to investigate what caused the problem.
Start with a problem your team has already solved. Compare the explanation with what you know and open the records, logs and code used to support it. Operate uses verification agents to check findings, but your engineers still need to review the answer and any suggested change.
You install Operate with Docker, connect an AI model and give it accounts that can read the systems it needs. The time depends on your access approvals and the systems involved. Our team can help with setup in a Managed engagement.
Your company gets 30 investigations a month, with unlimited Slack and Teams members. You can connect available tools, ask questions and receive explanations with evidence and suggested fixes. Follow-up messages in the same investigation are included.
If you use your own key or model, you pay that provider directly and Operate adds no token surcharge. If you use our AI service, usage costs $3.00 per million input tokens and $15.00 per million output tokens, in addition to the plan price. This applies to Free too. Unlimited investigations means questions started by people; bulk automated work and extra monitoring are agreed separately.
Pricing
Unlimited team members on every plan. No per-seat fees.
$0
per month
Connect your team
$99
per month
Investigate without limits
$299
per month
Monitor production
$499
per month
Build around your systems
Custom
Talk to us
Our engineers + Operate
Prices in USD, per company workspace. Follow-ups in the same investigation are included.
Unlimited covers investigations started by people. Bulk automated work and extra monitoring are agreed separately.
For open-source maintainers
A bug report describes what went wrong. Finding the cause still takes time. Link your public repository and use Operate to investigate its issues against the code.
For example, if an issue says a file export fails, Operate can examine the report and the export code to look for the cause. You can review what it found and the evidence behind the answer.
Add your public GitHub repository on the open-source page.
Operate reads the issue and relevant code to work out what may be causing the bug.
Check the findings and evidence. Public investigation pages can be linked from your issue discussion.
It could be an error alert, a customer question or a job that keeps failing. We’ll walk through it with you and show what Operate can check.
Book a demo