Convoy Apps

Long-running compute.

Two workload primitives on one pipeline. The platform builds, schedules, load-balances, and monitors the process. TLS, health checks, rollbacks, and log aggregation are handled at the infrastructure level.

ServicesLive
+
WorkersSoon
Join the waitlistWatch a deploy

LKR 500 welcome credit · No card required on free tier

01Workload02Deploy03Runtime04Recovery05Networking06Isolation07Scaling08Roadmap09FAQ
01Workload & source

Two workload types. One unified pipeline.

Pick a workload primitive, then point it at code or an image. Services answer HTTP and get a public URL with TLS. Workers run headless against queues, schedulers, or pollers. Both deploy through the same build, ship, and observe path.

1Workload
Available

Service

Long-running HTTP process. Public URL, TLS, health-check gate, horizontal autoscaling.

HTTPTLSPublic URLAutoscale
Coming soon

Worker

Headless background process. Queue consumers, schedulers, pollers, ETL jobs.

Non-HTTPConcurrencyRestart policy
2Source

GitHub source deploy

Authorize the Convoy GitHub App, then select any repository and branch.

acme/api-gateway
main
Auto-deploy on pushToggle off any time
On your next push
Trigger
push to main
Builder
Auto-detect
Image
private registry · last 10 kept
URL
api-gateway-a3f9c1.codeship.site
02Build and deploy

Five-stage pipeline, under sixty seconds.

Each deployment passes through validation, build, image push, rollout, and health check. Build logs stream in real time. Failed deploys are caught at the health-check gate, and the control plane exposes retry, cancel, and promote operations on every deployment.

api-gateway·acme-team
main·a3f9c1
ValidatingChecking source and limits
BuildingCreating your image
DeployingRolling out the change
StartingWaiting for health checks
LiveServing traffic
build.log
Try a control:

Auto-deploy on every push, or trigger from the dashboard.

Build logs stream live, no refresh, no tail.

Auto-rollback if a deploy fails health checks.

03Runtime telemetry

Continuous observability across every instance.

The runtime surfaces request rate, latency distribution, error rate, CPU, and memory utilization per instance without additional instrumentation. Deeper analysis is available through the integrated observability stack.

api-gateway·3 instances · eu-west-1
Live
Status
Healthy
Req / min
1,520
p95 latency
163ms
Error rate
0.02%
Requests / sec
63rps
CPU
53%
Memory
51%
Last 24 hoursavg 47 rps · peak 112 rps
24h ago12hnow
Activity streamstreaming
GET /api/orders · 142ms · 2000s ago
POST /api/checkout · 318ms · 2007s ago
deploy a3f9c1 promoted to live14s ago
GET /healthz · 4ms · 20021s ago
scaled 2 → 3 instances · cpu pressure28s ago
Logs, metrics, and traces

Structured log aggregation and metric collection, queryable in seconds, with configurable alert rules.

Per-tenant Grafana org

Each workspace gets an isolated Grafana organization scoped to its own data. SSO-bridged from the control plane.

04Deployment history

Automatic rollback on failed health checks.

Recent successful deployments remain promotable without a rebuild. When a release fails its health-check gate, the platform keeps traffic on the currently-serving revision and rolls back to the previous rollback target automatically.

Deployment ledger
Pinned releases are promotable, no rebuild required
LivePinned in registryEvictable
Rollback pipeline
Triggered on terminal deploy failure
Standing by
  1. 1
    Image probe
    currently-serving image still in registry
  2. 2
    Circuit breaker
    2 fails / 10m → disabled 1h
    1 / 2 in window
  3. 3
    Patch workload spec
    image ref rewritten to serving revision
  4. 4
    Atomic pointer swap
    currentDeploymentId → rollback target
typical recovery completes inside 60 seconds
05Networking

Managed TLS and custom domain routing.

Every deployment is assigned an HTTPS endpoint on the codeship.site domain. Custom domains are verified via CNAME, and TLS certificates are provisioned and renewed automatically.

Default URL

Every app gets a free codeship.site subdomain at deploy time

2Add custom domain

Point a CNAME and we verify DNS automatically

3TLS issued · live

Certificate provisioned and auto-renewed

https://api-gateway-a3f9c1-acme.codeship.site
200 OK · 47ms
{
  "service": "api-gateway",
  "version": "a3f9c1",
  "status": "healthy"
}

Live at deploy time. HTTPS, valid cert, ready to share.

HTTPS by default
Auto cert renewal
Multi-domain
HTTP→HTTPS redirect
06Storage · access control · isolation

Persistent state, scoped access, tenant isolation.

Production workloads require durable storage, role-based access, and hard multi-tenancy boundaries. These are infrastructure primitives, not optional add-ons.

Volumes

Block storage that persists across deploys.

Attach replicated volumes to any app. Data survives container restarts, redeployments, and node migrations.

postgres-data
20 GiB · replicated ×2
mounted
uploads
50 GiB · replicated ×2
mounted
redis-state
5 GiB · replicated ×2
mounted
Team

Role-based access per workspace.

Workspaces are the organizational boundary. Owner, Admin, and Developer roles control who can deploy, configure, and manage resources.

priyaOwner
marcusAdmin
julesDeveloper
ariDeveloper
Isolation

Hard multi-tenancy at the account level.

Cross-tenant traffic is blocked by default. Usage limits and encrypted per-workspace secret stores keep each account isolated.

Network policy
ingress denied across workspaces
Scoped secrets
encrypted at rest, per workspace
Resource quota
CPU + memory + storage limits
07AutoscalingComing soon

Horizontal scaling driven by real-time signals.

Define min and max instance counts. The platform monitors CPU, memory, and request throughput, and adjusts the replica count within seconds. Idle workloads will be able to scale to zero.

On the wayComing soon
Horizontal pod autoscaling

Instance count adjusts based on CPU, memory, and request-rate thresholds within seconds.

Scale to zero

Idle processes are terminated. The next inbound request triggers a cold start and resumes the workload.

Custom metric triggers

Scale on queue depth, p95 latency, or any metric published to the telemetry pipeline.

Sub-second cold start

Image layers are pre-cached across the node pool. Scale-from-zero latency stays under one second.

Currently, instance count is configured manually per app. Autoscaling is the next scheduled infrastructure primitive.
08Roadmap

Planned infrastructure primitives.

The current compute surface is intentionally scoped. These are the primitives under active development, in approximate shipping order.

Coming soon

Workers

Non-HTTP background processes with the same deployment pipeline. Configurable concurrency and restart policies.

Coming soon

Autoscaling

Horizontal pod autoscaling driven by CPU, memory, and custom metric thresholds.

Coming soon

Scale to zero

Idle workloads are terminated. Inbound traffic triggers automatic cold start.

Coming soon

Volume snapshots

Point-in-time snapshots with scheduled retention policies. Restore into a new volume in one operation.

09Common questions

Frequently asked questions.

Infrastructure layer

The compute runtime is built on production-grade open-source infrastructure with sensible defaults at every layer.

Isolation
tenant boundaries
Builds
source to image
Deploys
auditable rollout
Registry
controlled images
Storage
persistent state
Ingress
managed routing
TLS
auto certificates
Telemetry
metrics + logs
Get started

Deploy compute. Observe it run.

The free credit on every new workspace covers initial deployments. No payment method required on the free tier.

Join the waitlistTry the CLI
Free workspace credit·Usage-based billing in LKR·No fixed plans