Use Case · Agents

AI agent backend on the Software Lifecycle Cloud

An AI agent backend on KERNSTACK is the infrastructure pattern for letting a coding agent change a real system safely: the agent works on an agent/<id> branch in hosted git, its work deploys to an isolated agent-<id> environment, kernstack plan expresses the change as creates, updates, and deletes, and a person approves before production moves. The backend being one stack.json project is what makes the agent’s diff reviewable at all.

Agent SandboxesHosted GitFunctions (WASM)stack.json IaC
Talk to us Read Technical Documentation → Software Lifecycle Cloud →

The engineering bottleneck with assembled stacks

Teams who want an agent to open pull requests against a real stack discover the missing floor: the agent can edit code, but the system it edits is spread across consoles, so nobody can state what a proposed change does to infrastructure. The result is either a toy demo nobody trusts or an agent holding cloud keys nobody should have given it.

Architecture on the Software Lifecycle Cloud

Hosted git carries the repository; agent branches namespace the experiments. The agent edits code and stack.json together, runs kernstack plan --json to verify its own change set, and a reviewer checks the same plan. Environments multiply cheaply because a project is declared, not hand-assembled: the agent-<id> environment runs the branch’s stack for click-through review.

The plan is the agent’s report

An agent’s summary of its work is prose; a plan is evidence. Because the whole stack is declarative, the plan enumerates consequences — this function changes, this table is created, nothing is deleted — in a form both the agent (as JSON) and the reviewer (as a diff) can check against the original request. Destructive entries stop the line until a human passes --confirm-destructive.

  • Agent branches deploy to their own environments, so review happens at a live URL
  • A deploy gate can require review-approved source (min_source_assurance), making approval a platform property
  • Keeping agent work is one step: move it to feature/, which kernstack use offers when leaving an agent branch

What the agent never holds

The agent holds a branch, not your infrastructure. It cannot quietly widen its own access, because access inside a project is structural — resources belong to the project declared in stack.json — and production credentials are never the agent’s to present. When a demo is discarded, its branch and environment go with it.

FAQ

Frequently asked questions

None in particular. Cursor, Claude Code, Codex, Pi, and the project’s own @agent all produce the same artifacts — branches and plans. KERNSTACK governs what happens to those artifacts; model choice stays yours and can change per task.

Read kernstack plan: it lists creates, updates, and deletes against the live stack before anything applies. Combined with the code diff on the branch, that is the complete change — there is no console where a third, unreviewed part could hide.

Only if you dismantle the gates on purpose. The sane configuration keeps agent work on agent branches and agent environments, requires review-approved source for production deploys, and reserves --confirm-destructive for a human hand.

Connected in the Project Graph

Related capabilities, architectures, and comparisons

Category PillarThe Software Lifecycle CloudWhy people and agents build, deploy, and operate on a single vertically integrated stack.Service · AgentsAgent sandboxesIsolated sandboxes for coding agentsService · SourceHosted GitRepository hosting inside the projectService · ComputeFunctionsGo WebAssembly HTTP handlersService · Manifeststack.jsonOne manifest for the whole projectIndustry SolutionHealthTech Cloud SolutionsHealthTech products on KERNSTACK hold care-adjacent data — patient documents, encounter records, consent forms — in one declared project in a European zone on named infrastructure. The needs that dominate are unglamorous and absolute: know exactly where every document lives, who touched it, how consent was captured, and how a record can be frozen the day a legal hold arrives.Use Case · ProductExperiment assignmentAssignment is a pure function of the subject key and the experiment’s salt — stable across requests, no coordination needed. The function records the assignment row once and serves the variant; outcome handlers were already writing events, which now simply carry the experiment id the assignment returned. Analysis reads assignments and outcomes from neighbouring tables with the same keys.Use Case · AgenciesAgency client deliveryThe studio’s delivery shape is a template stack.json: gateway, functions or workers in the client’s stack languages, tables, buckets, and apps under one target per client, on European infrastructure the studio can name in the pitch. Previews run the same declaration; the client’s approval happens in the bound conversation with a live link. Every client’s meter is separate by construction — the studio sees delivery cost per client without allocation spreadsheets.

Ship your stack on the Software Lifecycle Cloud.

Closed preview on dedicated European infrastructure. Bring a repository or start from a blueprint.

Talk to us See pricing & credits →