Skip to content

The environment is the center of delivery

Pipelines know how to run steps. Massdriver knows what you’ve deployed, how it connects, and what changes are allowed.

From pipeline-driven delivery to environment-modeled delivery.

Traditional delivery
Delivery knowledge tends to spread across workflows, infrastructure code, state, configuration, credentials, and environment-specific logic. The next change has to reconstruct enough of that context to know what it is changing.
billing-apiorders-apiwebdeploy.ymlaws credentialsvpc + subnetspostgres urlprod approvalstate backendterraform applydeploy.ymlaws credentialsvpc + subnetspostgres urlprod approvalstate backendterraform applydeploy.ymlaws credentialsvpc + subnetspostgres urlprod approvalstate backendterraform applyyour cloud
Massdriver
The rules, relationships, and credentials live in one model. Every request goes through it.
billing-apiorders-apiwebENVIRONMENTprod approvalpostgresqueueWORKFLOWSTerraform · OpenTofu · HelmAPPROVED EXECUTIONyour cloud
01

CI builds, tests, and packages

Keep the pipeline for what it is good at: compiling, testing, and pushing an image or zip. Its last step hands the artifact to Massdriver.

.github/workflows/ship.ymlConcept syntax
A CI workflow that builds and tests, then hands the deploy to Massdriver
- run: npm test
- run: docker build -t $IMAGE . && docker push $IMAGE
- run: mass deploy billing-api --env dev --image $IMAGE
02

You, or an agent, state the deploy

The request names the app, the artifact, and the target environment. It does not describe how to reach the cloud.

intentConcept syntax
The developer states the deploy: which app, which image, which environment
$ mass deploy billing-api --env dev --image ghcr.io/acme/billing:7f3a1c
03

Massdriver resolves it against the environment

Cloud credentials, networks, and connections are registered once, along with what is using each of them. The deploy picks up the right ones for that environment, so nobody copies keys, VPC IDs, or connection strings between pipelines. The delivery workflow does not have to reconstruct the environment from variables, outputs, and disconnected configuration.

shop-devThin UX
Massdriver resolves the request against the environment: runtime, credentials, network, and dependencies
resolving billing-api in shop-dev
runtime lambda@1.2.8 platform team
credential aws-iam-role env default
network aws-vpc/shop-dev env default
database postgres linked
queue orders-queue linked
04

Typed dependencies decide the wiring

Resources emit typed outputs and apps declare typed inputs. A postgresql output satisfies a postgresql input, so connection details, IAM, and network rules flow in without copy and paste.

typed connectionToday
A postgresql resource satisfies billing-api’s database dependency and its fields are wired in
postgres emits postgresql
billing-api.database expects postgresql ✓
DATABASE_URL ← postgres.auth.hostname, port, …
IAM ← postgres.policies.read_write
SECURITY GROUP ← postgres.network.ingress
05

Approved infrastructure runs underneath

Each change is planned, checked against policy, approved if the environment requires it, and executed with Terraform, OpenTofu, or Helm in a short-lived worker that holds the credentials.

deployment d-7f3aToday
The deployment plans, passes policy, runs OpenTofu in an ephemeral worker, and stores state
✓ plan postgres create, billing-api update
✓ policy dev-default (3 rules)
✓ approval not required in dev
✓ run opentofu 1.9 · ephemeral worker · 3m12s
✓ state stored, outputs published as resources
06

The environment stays understandable

The result of the deployment remains part of the same model: what was deployed, which infrastructure it uses, what it depends on, what depends on it, and what controls apply.

Humans, automation, and agents can start the next change with that context already available.

postgresqlsqs-queuehttp-apibilling-apiapp · lambdapostgresaws-rds-postgres@1.4queueaws-sqs@0.9payments-apiapp · ecs
07

The next change starts from the same model

Upgrades, new dependencies, and promotions to prod all operate on the environment. Impact analysis comes from the graph, and policy applies the same way every time.

agent sessionThin UX
A coding agent asks what depends on orders-db, then proposes an upgrade that goes through plan, policy, and approval
agent> what depends on orders-db in prod?
orders-db aws-rds-postgres@1.3
├── billing-api
├── worker
└── reporting-job
agent> propose upgrading orders-db to 1.4
plan 2 changes, 0 destroys
impact 3 dependents, connection params unchanged
policy ✓ passed (prod-db-change)
approval required from @platform
→ proposal #418 waiting for review

Try it on a real app