# Roadmap

This roadmap describes ideas Keel is exploring, not committed features or release dates. Items are listed in approximate priority order, with current behavior noted where useful.

Recently graduated off this page and into the product: [static sites & CloudFront](./guide/static-sites), [buildpacks](./guide/deployments#buildpacks), [scheduled tasks](./guide/scheduled-tasks), [cost estimates & budgets](./guide/cost), queue-driven scale-to-zero, [internal services behind a login](./editions/pro#internal-only-services-behind-a-login), [database tunnels and shells](./database-credentials#reaching-the-database-at-all), [EC2-backed capacity](./guide/configuration#placement-and-capacity), and [operator access through the control plane](./guide/operators).

## Agentic AI harness

An assistant inside Keel that can use `keel.yml`, deployment history, logs, and live AWS state to help configure and operate an application.

What that looks like in practice:

- **Plan changes in plain language.** A request such as “add a smaller staging environment in us-west-2” becomes a reviewable `keel.yml` diff and the commands needed to apply it.
- **Draft configuration from a repository.** Detect services, health checks, release commands, and likely secrets, then propose an initial `keel.yml`.
- **Diagnose failed deployments.** Correlate preflight output, CodeBuild logs, stopped-task reasons, and CloudWatch logs into a cause and suggested fix.
- **Query live operations.** Answer questions about recent changes, resource use, and application behavior from live state and metrics.
- **Recommend autoscaling settings.** Use observed service behavior to propose a configuration diff.

The proposed design keeps changes reviewable and requires approval before execution. Actions would use Keel's existing IAM levels or operator capabilities, with the control plane recording authorization decisions.

## App Library

Curated Keel manifests for pre-built images such as Grafana, Metabase, and PostHog, with a focused configuration surface and standard service operations.

## Plugins

A plugin interface for adjacent workflows that do not belong in the core binary. Database tunneling and `keel db shell` have shipped as core commands, so the remaining design question is the extension boundary rather than that first use case.

## Database performance insights

RDS Performance Insights in the dashboard, including query latency, long-running queries, and comparisons across time ranges.

## Guided database upgrades

A guided workflow for database upgrades that explains compatibility, expected downtime, minor-version policy, and the proposed plan before applying it.

## Read replicas & multi-node databases

`multi_az` already provides a failover standby. This item would add readable replicas and reader endpoints for scaling read traffic.

## Deployment strategies

Today every deploy is an ECS rolling deployment with a circuit breaker — a bad revision rolls back instead of looping, and the release command gates promotion. Making the strategy configurable — blue/green, canary percentages, bake time — is the roadmap item.

## Scheduled autoscaling

CPU-, memory-, request-, and SQS backlog-based target tracking shipped (`target_queue` also enables scale-to-zero). Still planned: time-based scaling schedules, and suggestions derived from your actual usage — the latter is part of the [agentic AI harness](#agentic-ai-harness).