A team of AI agents that builds, rewrites, maintains and operates your systems end to end. Connected read-only on day one. A person approves every change.
GitHub, Azure DevOps or GitLab · Azure or any Kubernetes cluster · nothing changes without your yes
Most tools cover one phase. DevCue is one team that stays with the system from the first plan through years of operation.
Describe the change in a sentence. Planning studies the system, Design writes the impact note, Test writes the cases, Development opens reviewed pull requests, Docs updates the wiki.
Modernize without a leap of faith: an inventory from code and the live schema, a proposal that prefers keeping and simplifying, a parity baseline, reviewable slices, a staged cutover with rollback.
Security scans, Microsoft Defender findings, versions near end of support, cloud cost and savings, pull request reviews, tests and documentation, all on a schedule, all as work items with evidence.
Addresses and clusters watched around the clock. A failure becomes an incident with a likely cause, a work item in your tracker, and a proposed fix waiting for one tap in the morning.
Not a code generator with tests bolted on. One platform that maps, plans, builds, tests, documents, watches, secures and sells.
Connect code and cloud read-only. Every app, how it is built and deployed, its dependencies, tests, resources and the gaps between them, minutes after connecting.
Study, Enhancement, Bug fix, Security. The agent reads the code, inspects the live schema, observes the app in a browser, and writes down what it could not cover.
Suites found and written in your code’s own stack, UI included, run in an isolated test machine, results recorded against the plan that asked for them.
Overview, workflows, screens, rules, data model, architecture, runbook: written by plans into the wiki and updated when a change lands.
Addresses checked, clusters watched. A repeated failure becomes an incident with a likely cause and a next step, and a work item in your tracker.
Code and image scans, software bills of materials, Microsoft Defender findings, versions near end of support. Billing, cluster usage and savings with evidence.
One queue for problems, fixes, incidents and savings, mirrored to Jira, Azure DevOps or GitHub and to Slack or Teams. Releases queued only when your policy allows.
Five templates and a builder. Tools from your connections, triggers on a schedule, an event, a work item or another agent. Try it before Publish. A daily spend cap.
Campaigns on Google, Meta and TikTok built paused for review, ad text within each platform’s limits, results watched, leads scored and sent to HubSpot.
The same team, pointed at the job in front of you.
Describe the product. The team plans it, writes it, tests it, deploys it to your cloud, and then keeps running it. HeyCue’s 67 business apps were built this way.
A project ends, the system does not. Connect it and the team watches it from day one: incidents, releases, security, cost.
Monitoring, incidents and a first fix while people sleep. Mornings start with proposals to approve, not pages.
Describe the enhancement. Study, design note, test cases, work items, reviewed pull requests, documentation. You approve four times.
Paste the error. The agent reproduces it, stops before saving anything, and writes the fix as a pull request with its tests.
Inventory from code and schema, a proposal that keeps what works, a parity baseline, slices as pull requests, a staged cutover with rollback.
One Study writes the overview, workflows, screens, rules, data model and runbook, and says what it could not cover and why.
Scans and Defender findings on a schedule, not only on the loud systems. Findings become work with evidence and get closed.
What each system costs, where the money goes, idle capacity named, savings proposed as work items you can approve.
Single copies, missing resource requests, full clusters, Kubernetes versions near end of support, with the change written when you ask.
One engineer manages a department of agents per client: approves its proposals, sets its rules, answers its questions. More clients per head, with an audit trail.
Campaigns built paused, landing pages checked for tags, results watched, every lead scored and in HubSpot within minutes.
A coordinator hands out the work between four departments. Your people approve what changes.
Planning · Design · Development · Review · Test · Docs
Deploy · Monitor · Incident response · Infrastructure
Security · Cost
Marketing · Creative · Ad results · Sales
Hire more: your own agents sit in the same departments, with five templates, Try it before Publish, a daily spend cap and the same rules.
Watch, decide, act, ask, learn. Every night. Here is one, as the Activity page would show it.
An address check fails twice. A problem opens in Work.
Opens the incident, reads the logs, names the likely cause.
A fix as a pull request. The test suite runs against it.
Reviewed and ready. One tap in the morning: approve the merge.
The release goes out, the runbook is updated, Monitor confirms.
Agents that only read keep running even when the team is stopped. Anything that writes becomes a pull request, a draft or a proposal, and waits for a person.
Governance is not a module. The write rule, approvals and spend limits sit in the runtime every agent runs through.
Nothing changes in your systems until a person says yes. A fix starts on its own only where your fix policy allows it.
Every action, by every agent and person, in one log you can export.
Owner, admin, member, viewer. Private systems stay private to the people named on them.
Dollars and runs per day on every agent you add. It stops at the cap.
One button. Agents that only read keep watching; everything else halts.
Code, clouds and clusters are connected read-only. A person widens it, per system. Secrets are stored once and never come back in an answer.
InvoiceCue and HeyCue run on DevCue every day.
Our invoicing product: .NET services, Azure DevOps, Azure SQL, two Kubernetes clusters. 7 apps and 17 repositories mapped. One Study wrote 14 wiki pages and 74 test cases. 9 production addresses watched with 18,000+ checks, 102 work items and 22 fix proposals. 2,257 security scans, 421 findings triaged, cost tracked every day.
Our small-business product: a phone agent, a website and a booking or shop app per business. 67 business sites and apps built from conversations, each through a compile, test and repair loop. A running app changed by sentence in Studio. Security, cost, tests, monitoring and incidents on HeyCue’s services with the same team and the same rules.
InvoiceCue counts read from its DevCue workspace on 26 October 2026. HeyCue count as of October 2026.
A 30-day pilot on one system, then a monthly price per workspace. Tell us what you run and we quote it the same day.
What people ask before they connect a system.
No. Agents read first. Any change arrives as a pull request, a draft or a proposal and waits for a person. A fix starts on its own only where your fix policy says it may, and even then it is a pull request, not a deploy.
Read-only access to your code on GitHub, Azure DevOps or GitLab, and to where it runs: Azure or any Kubernetes cluster. Trackers, chat, your CRM and ad accounts are optional. A person widens access per system when they want to.
Any. DevCue reads the repository, the pipelines and the manifests as they are, and runs tests in the languages your code already uses. .NET, Go, Node, Python and Java systems are in workspaces today.
DevCue is a hosted workspace; your code, databases and clusters stay where they are. Connection strings and keys are stored once, sealed, and never come back in an answer. Agent instructions cannot hold a secret.
Owners, admins, members and viewers, set per workspace and per system. Private systems are visible only to the people named on them. Every approval and every agent action is in the audit log, which you can export.
It grew up. The engine that generated whole apps from a description is now the Build department of the team, next to the agents that test, secure, monitor and operate what gets built.
A 30-day pilot on one system, then a monthly price per workspace. Tell us what you run and we quote it the same day.
See the map of your system in minutes. Nothing changes until you say yes.
Read-only on day one · A person approves every change · Stop all, one button